Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
HOW WELL IS THE RUSSIAN WHEAT MARKET FUNCTIONING?
A COMPARISON WITH THE CORN MARKET IN THE USA
Miranda Svanidze, Linde Götz
Leibniz Institute of Agricultural Development in Transition Economies
(IAMO), Halle (Saale), Germany
2017
Copyright 2017 by authors. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies.
Paper prepared for presentation at the 57th annual conference of the
GEWISOLA (German Association of Agricultural Economists)
and the 27th annual conference of the ÖGA (Austrian Society of Economics)
“Bridging the Gap between Resource Efficiency and Society’s Expectations in
the Agricultural and Food Economy”
Munich, Germany, September 13 – 15, 2017
HOW WELL IS THE RUSSIAN WHEAT MARKET FUNCTIONING?
A COMPARISON WITH THE CORN MARKET IN THE USA
Abstract
Given Russia’s leading position in the world wheat trade, how well its grain markets function
becomes very important question to evaluate the state of future global food security. We use a
threshold vector error correction model to explicitly account for the influence of trade costs and
distance on price relationships in the grain markets of Russia and the USA. In addition, we study
impact of market characteristics on regional wheat market integration. Empirical evaluation
shows that distance between markets, interregional trade flows, export orientation, export tax and
export ban all have a significant impact on the magnitude of wheat market integration.
Keywords
regional market integration, threshold vector error correction model, grain markets, Russia, USA,
export ban.
1 Introduction
In recent years Russia has advanced from a grain importing country to one of the primary grain
exporting countries. In 2016/17 Russia is forecasted to become the largest wheat exporter in the
world (Interfax 2016).
Russia could further boost its grain production by increasing production efficiency and also by re-
cultivating formerly abandoned agricultural land. According to OECD/FAO (2012) global grain
production needs to increase by 30% to satisfy global demand for cereals which will reach 3
billion tons by 2050. Russia could play a large role for future global food security (Lioubimtseva
and Henebry, 2012). This requires not only Russia’s large additional grain production potential to
be mobilized but also that grain markets are functioning well enabling that the grain exporting
potential is mobilized as well.
This study aims to address the research question how well the Russian grain market is
functioning, a question which has not been addressed in the literature before. Following a price
transmission approach we are focusing on the primary grain producing regions and investigate the
integration of the regional grain markets. To what degree and how fast are price shocks in one
region transmitted to the other regions?
This is an important question given that the Russian grain market is characterized by strong
production volatility resulting from extreme weather events which are expected to increase with
climate change. Favourable production conditions and thus relatively high yields can be observed
in some regions but relatively low yields in other regions at the same time. Therefore,
interregional grain trade is of high importance to equilibrate grain supply and demand within
Russia. Nonetheless, grain market transport and storage infrastructure is deficient in several
regions and price peaks are repeatedly observed on regional markets, exceeding even the world
market price.
In a well-functioning, efficient market, with a well-developed transport and storage infrastructure,
regional prices differ at most by the costs of trade between those regions. Also, price shocks in
one region are quickly transmitted to the other regions inducing interregional trade flows when
price differences exceed trade costs (Fackler and Goodwin, 2001). Thus, an efficient market could
also contribute to cushioning price increasing effects of regional harvest shortfalls and prevent
that prices increase beyond the world market price.
However, Russia has a history of restricting the exports of wheat to the world market when
domestic wheat prices peak. As our second question, we investigate the effects of the wheat
2
export ban 2010/11 on regional price relationships to shed further light on the domestic price
effects of export controls. This is an addition to Götz et al. 2013, 2016 which focus on the export
controls’ effects on the integration in the world market.
We address both research questions in a price transmission framework. We apply a threshold
vector error-correction model (TVECM) to explicitly account for the influence of distance and use
a Bayesian estimator suggested by Greb et al. (2013) as an alternative to the conventional
maximum likelihood approach (Hansen and Seo, 2002; Lo and Zivot, 2001).
Highly integrated markets characterized by strong price relationships with fast transmission of
price changes between the regions are usually interpreted as evidence for well-functioning
markets. However, the Russian wheat market is characterized by extremely large distances of up
to 4000 km which certainly negatively affects market integration.
To assess how well the Russian market is functioning we conduct a comparative price
transmission analysis for the corn market of the USA which is also characterized by large
distances, strong variation in regional production and high interregional trade flows. We assume
that the corn market of the USA is one of the most efficient grain markets in the world
characterized by well-developed transport and storage infrastructure and high market
transparency, serving as a benchmark for the Russian wheat market in this study.
The remainder of this paper is organized as follows. In the next section, we discuss market
conditions and the consequences of the export ban 2010/11 for wheat trade in Russia. This is
followed by the review of major literature sources. The detailed presentation of econometric
model is given in the section on methodology and estimation. Data section 5 focuses on the
properties and preliminary assessment of time series used in analysis. In the results section6, we
discuss outcomes of model estimation. The results for the Russian wheat market are compared to
the results for the corn market of the USA presented in section 7. In the final section 8,
concluding remarks are given.
2 Characteristics of the Russian wheat markets
2.1 Regional wheat production and harvest shortfall
Wheat production in Russia is concentrated on a limited, yet spatially protracted area. Two large
regional production clusters emerge depending on both the type of culture and the area of
cultivation. The winter wheat cluster covers Southwest of Russia stretching from the Black Sea to
Volga. Yields in this area amount to 3 tons per hectare on average (2006-2010). The spring wheat
cluster spans over Urals and West Siberia. In contrast to winter wheat, spring wheat is much less
productive with yields amounting to 1.7 tons per hectare on average (2006-2010).
During the last decade, 2010 marks as the year when Russian grain markets witnessed
exeptionally low production. Drought in 2010 affected the key crop growing areas in Russia
(Figure 1), whcih resulted in the unusually low harvest and export-respricting policies by the
government.
Figure 1. Map of crop-growing regions affected by drought in 2010
3
Source: Own illustration.
In general, the size of wheat production in the major grain production regions differs strongly. As
can be seen from Figure 2, wheat production is highest in North Caucasus, with an annual
production varying between 12 and 22 million tons in the period 2005-2013. This is followed by
Volga and West Siberia (with wheat production varying between 4 and 11 million tons in each
region), Black Earth (between 3 to 9 million tons), Urals (between 2 to7 million tons) and Central
(2-4 million).
Figure 2. Regional wheat production development in Russia
Source: Götz et al. 2016
In addition, the variation of wheat production within a region is also extremely high. Table 1
gives regional wheat production as the average of wheat production in the previous three years. It
becomes evident that for example in the Volga region, wheat production varied between 34% and
143% in the marketing years 2005/6 to 2012/13. Weather conditions are a key determinant of the
quantity of wheat production. Due to large distances, the production regions are affected by
different climatic and weather conditions.
Table 1: Regional wheat production developments Russia (2005-2013), in % of the average
of previous 3 years
20000
28000
36000
44000
52000
60000
0
5000
10000
15000
20000
25000
2005 2006 2007 2008 2009 2010 2011 2012 2013
10
00
t
10
00
t
North Caucasus West Siberia Volga Black Earth
Ural Central Total
4
2005/6 2006/07 2007/08 2008/9 2009/10 2010/11 2011/12 2012/13
North Caucasus 126 112 98 139 107 104 108 68
Central 109 100 117 160 159 81 79 97
Black Earth 120 97 117 187 136 46 86 96
Volga 112 103 98 143 109 34 81 84
West Siberia 93 98 114 99 140 86 96 46
Ural 91 118 125 117 87 38 144 61 Source: Götz et al. 2016
This implies that favorable production conditions and thus relatively high yields might be
observed in some regions but relatively low yields in other regions at the same time. For example,
grain production was even by 4% above average in North Caucasus in the marketing year
2010/11, whereas the regions Volga, Urals and Black Earth were severely hit by the drought with
grain production 66%, 62% and 54% below average, respectively.
2.2 Regional wheat trade and transportation costs
The Russian wheat producing regions can be classified into surplus and deficit areas. The former
includes North Caucasus, Black Earth, Volga, West Siberia and Urals, which usually supply their
production excess to other markets. Central region with Moscow is the primary wheat deficit
region, which heavily depends on external supplies. Central mostly imports wheat domestically
from the regions Black Earth, Volga, Urals and West Siberia. By contrast, North Caucasus
supplies primarily to the world markets, while its role in the domestic trade is rather limited. Due
to the presence of high-capacity sea terminals, North Caucasus also serves as a gate-market for
the other grain producing regions, particularly Volga and Black Earth, to export wheat to the
world market. Differing, Urals and West Siberia are far away from not only the world market,
with the distance to the Black Sea ports amounting up to 4000 km, but also the grain consumption
regions within Russia. In particular, Moscow is about 2000-3000 km apart. Due to outdated and
insufficient transport infrastructure, Urals and West Siberia are not well connected neither world
market nor the consumption centers.
However, due to the large variation in grain production, the size of trade flows between surplus
and deficit regions may vary strongly.
Trains and trucks are the two primary means of wheat transportation in Russia. Trains are mostly
used when the transport distance between regions exceeds 1000 kilometers, while trucks are often
preferred on shorter routes.
As production areas cover large territory, the influence of transport infrastructure is crucial for the
distribution of wheat. The quality of transport infrastructure strongly differs between regions. For
instance, the density of the railway network is highest in the European part of Russia, whereas it is
much lower in Urals and West Siberia. It is reported that excessive crops are often difficult to
transport beyond West Siberia as the only railway track connecting the area to the rest of the
country has low throughput capacity and is shared by many other industries (Scherbanin, 2012).
In addition, grain traders regularly complain that the number of grain wagons in peak seasons
does not suffice (Gonenko, 2011).
The estimated values of the railway delivery fees for selected market pairs in 2010 are presented
in Table 2. It should be pointed out that the delivery fee captures only parts of the full transport
costs. Other expenses may include storage fees, transportation between the railway station and the
grain processing facility, insurance premium etc. The share of the delivery fee in the total
transport costs varies significantly amounting to 30% to 70% of transport costs.
Table 2. Costs of wheat transportation between the selected locations in 2010
Pair of
markets
Station of
origin
Station of
destination
Distance
(km)
Delivery fee
(RUB/ton)
Delivery fee
(USD/ton)
North Caucasus-Black Earth Kavkaz Voronezj 870 781 26
5
North Caucasus-Central Kavkaz Moscow 1300 1165 39
North Caucasus-Volga Kavkaz Kazan' 1708 1328 45
Volga-Central Kazan' Moscow 812 752 25
Urals-Central Kurgan Moscow 2037 1498 50
West Siberia-North Caucasus Novosibirsk Kavkaz 3800 2576 86
West Siberia-Central Novosibirsk Moscow 3350 2147 72
Note: Delivery fee is recognized as a charge due to be paid for the rent of one wagon (measured in RUB/ton). The
value of delivery fee is estimated using an online calculator provided by the Russian railways on August 06, 2010
when trade was freely possible. These estimates correspond to the amount of wheat which is simultaneously
transferred in a group of 100 wagons. Therefore, they may slightly vary if the actual number of wagons included in
the group differs.
Source: Own illustration, data: Rosstat 2015.
3 Literature review
This paper adds to the strand of literature focusing on spatial price relations between regional
agricultural markets.
Goodwin and Piggott (2001) first introduced threshold co-integration in the spatial price
transmission literature. They analyse spatial price links between regional corn and soybean
markets in North Carolina using a two-regime threshold autoregressive (TAR) model. They find
that thresholds are proportionally related to transaction costs, which increase with distance
between the markets. Their study confirms the presence of non-linear adjustment of prices to
deviations from the long-run price equilibrium between two locations. In particular, price
adjustment is hardly confirmed if regional price differences are smaller than transaction costs. On
the contrary, large price differentials induce adjustment of regional prices to their price
equilibrium, which increases with proximity of the markets. Additionally, the authors utilize a
three-regime threshold vector error-correction model (TVECM) to account for changes in the
direction of trade flows. However, model results do not find evidence that a reversal in trade
direction alters the speed of price adjustments to its spatial price equilibrium.
Several methods have been developed to correctly identify the optimal threshold parameter. Chan
(1993) offers the method of threshold selection that gained recognition in the context of the TAR
model. According to this approach, the optimal threshold is to be chosen from the set of residuals
retrieved from the long-run equilibrium regression. The residuals are sorted using results of sum
of squared errors (SSE), and the residual with the lowest SSE is selected as a threshold. Hansen
and Seo (2002) use values of error-correction terms (ECTs) to determine possible threshold
adjustment in a two-regime TVECM. They pair ECTs with corresponding values of the co-
integrating vector to construct a two-dimensional grid and then estimate this grid with maximum
likelihood. The pair that yields the lowest value of the concentrated likelihood function is
determined to contain the optimal threshold parameter. These procedures are criticized for the
reliance on an arbitrarily chosen trimming parameter which is used to ensure that the model
parameters of each regime are estimated based on a minimum number of observations. According
to Greb et al. (2014) the selection of the trimming parameter might lead to the exclusion of the
true value of a threshold from the threshold parameter space and, as a consequence, to unreliable
threshold values and model parameter estimates.
Balcombe et al. (2007) offer an alternative framework to estimate the parameters of generalized
threshold error-correction model on the basis of classic Bayesian theory. They apply this model to
monthly wheat, maize and soya prices for the United States, Argentina and Brazil. Results suggest
that the new method is capable of addressing the problem of identification of model parameters
that often pertains to the maximum likelihood approach. This problem results from the jagged
nature of the maximum likelihood function implying that the function cannot be evaluated using
traditional differentiation methods. By contrast, classic Bayesian analysis offers special
computational algorithms which allow estimating the parameters without using the irregular
maximum likelihood function.
6
Greb et al. (2014) use a methodologically similar empirical Bayesian paradigm to develop a
threshold estimator in the context of generalized threshold models. However, in comparison to the
Bayesian approach followed in Balcombe et al. (2007), they tend to reduce the application of so-
called non-informative priors. According to Greb et al. (2014), certain prior values should be
assigned to the model parameters to make the estimation procedure possible, but in the absence of
any preliminary information this assignment becomes rather arbitrary and may influence the final
estimates. To avoid this outcome, they start with selected priors obtained from maximum
likelihood estimation. Additionally, the empirical Bayesian analysis requires no trimming
parameter to achieve the desired distribution of observations across regimes. Greb et al. (2013)
exploit this approach and compare it to the maximum likelihood procedure to revisit the study of
Goodwin and Piggott (2001). Applying three-regime TVECM, they conclude that the Bayesian
estimator identifies larger thresholds and wider inaction bands compared to the maximum
likelihood counterpart. Moreover, they also find more evidence in support of asymmetric
adjustment that takes place, potentially, due to changes in the direction of trade.
Our study also contributes to the growing price transmission literature on the domestic price
effects of export controls. The effects of wheat export controls in Russia were previously
addressed within a price transmission approach by Götz et al. (2016) and Götz et al. (2013). Both
studies focus on the relationship between the world market price and the domestic prices in order
to identify the price dampening effect of the export controls. Götz et al. (2013) investigate
domestic price effects of the export tax in Russia during 2007/8 within a MSECM approach. They
find compared to Ukraine a rather low price dampening effect amounting to 25%. Results of Götz
et al. (2016) suggest a strong heterogeneity of the price dampening effect of the wheat export ban
2010/11 in Russia, varying between 67% and 35% in the major grain producing regions.
Differing, this study investigates how the export ban 2010/11 impacts price relationships between
the grain producing regions of Russia themselves. A further novelty of our approach is that we use
a TVECM in order to capture the possible effects of the export ban on trade costs. Also, we are
supplementing the regional price data with interregional trade flow data to facilitate interpretation
of our model results.
A regional perspective is also followed by Baylis et al. (2014) which investigate the export ban
for wheat and rice implemented in India 2007-2011. They take into account integration between
the world and domestic markets, but also explicitly focus on price relations between the regions of
India. The analysis is based on regional price data for producing, consuming and port markets and
the world market price. Using a linear VECM and a TVECM, they investigate cointegration and
integration for the time period when trade was freely possible and compare it to when the export
ban was implemented. They find for rice all port markets integrated with the world market during
the export ban period as well as when trade is freely possible. Though, no cointegration of the port
markets and the world market for wheat is observed during the export ban. However, more
domestic market price pairs are integrated during the export ban for rice but less for wheat, when
compared to the free trade regime.
4 Methodological framework and data properties
4.1 Measurement of market integration
Regionally integrated markets are related through a long-run equilibrium parity, which we
characterize by long-run price transmission elasticities estimated in the cointegration equation.
Price transmission elasticities characterize how strongly are price shocks transmitted from one
region to another. Long-run price equilibrium is given as:
𝑃𝑡1 = α + β𝑃𝑡
2 + 𝜀𝑡 (1)
where 𝑃𝑡1 and 𝑃𝑡
2 are natural logarithm of prices at time t for every regional market pair, ε𝑡 denotes stationary disturbance term. α denotes intercept and β is coefficient of the long-run price
7
transmission elasticity, characterizing the magnitude of the transmission of price shocks from one
market to another. Regression equation is estimated by the ordinary least squares method.
Usually, market prices are tend to diverge from the long-run equilibrium parity from time to time.
Threshold vector error correction model (TVECM) is designed to examine how fast prices
converge back to the equilibrium in the short-run. We adopt a non-linear 3-regime TVECM with
2 thresholds developed by Greb et al. (2013) also to account for the influence of trade costs,
which are highly relevant to the Russian wheat market.
A three-regime TVECM is illustrated in equation (2). The vector of dependent variables ∆𝑃𝑡 =(∆𝑃𝑡1, ∆𝑃𝑡
2) denotes the difference between prices in periods 𝑡 and 𝑡 − 1 for both markets in
question. As the independent variables, 𝜀𝑡−1, error correction term, or alternatively, lagged
residuals from equation (1) is taken to represent the price deviation from the long-run price
equilibrium. Additionally, ∑ ∆𝑃𝑡−𝑚𝑀𝑚=1 term is the sum of price differences lagged by period m to
correct residual correlation, and 𝜔𝑡 denotes a white-noise process with expected value 𝐸(𝜔𝑡) = 0 and covariance matrix 𝐶𝑜𝑣(𝜔𝑡) = Ω ∈ (ℝ
+)2×2.
∆𝑃𝑡 =
�
𝜌1𝜀𝑡−1 + Θ1𝑚∆𝑃𝑡−𝑚 + 𝜔𝑡
𝑀
𝑚=1, 𝑖𝑓 𝜀𝑡−1 ≤ 𝜏1 (𝐿𝑜𝑤𝑒𝑟)
𝜌2𝜀𝑡−1 + Θ2𝑚∆𝑃𝑡−𝑚 + 𝜔𝑡𝑀
𝑚=1, 𝑖𝑓 𝜏1 < 𝜀𝑡−1 ≤ 𝜏2 (𝑀𝑖𝑑𝑑𝑙𝑒 )
𝜌3𝜀𝑡−1 + Θ3𝑚∆𝑃𝑡−𝑚 + 𝜔𝑡𝑀
𝑚=1, 𝑖𝑓 𝜏2 < 𝜀𝑡−1 (𝑈𝑝𝑝𝑒𝑟)
(2)
The short-run dynamics are characterized by the speed of adjustment parameter (𝜌𝑘) and the
coefficients of the price differences (Θ𝑘𝑚) lagged by m-periods with k referring to a regime. All
parameters may vary by regime with k=1 … 3.
Price observations are attributed to a certain regime depending on the size of the ECT. The 3-
regime TVECM is based on the assumption that two thresholds exist corresponding to the costs of
trade in both directions, i.e. from one market to the other and vice versa. Price observations for
which the ECT is smaller than threshold 𝜏1 are attributed to the lower regime, whereas price
observations with an ECT larger than threshold 𝜏2 are assigned to the upper regime 3. The
threshold is considered a proxy for transaction costs of wheat trade between the two respective
markets. If the ECT is of the size smaller than threshold 𝜏2 but larger than threshold 𝜏1, the
observations are allocated to the middle regime. Within this regime, the difference between the
prices of two regions are smaller than transaction costs of trade.
The speed of adjustment refers to the time period required by the price of a certain market to
correct a deviation from the long-run equilibrium between the two markets. The speed of
adjustment may differ between the regimes. Prices in two spatially separated markets may be
related by trade arbitrage only if the price differences are at least as high as trade costs. This is
given for price observations which are attributed to the upper and lower regime in a 3-regime
TVECM. However, prices may be related but at a lower degree even if the price differences are
smaller than transaction costs, corresponding to the middle regime in a 3-regime TVECM, via
information flows or third markets (Stephens et al., 2012).
There are several conditions that should be satisfied to ensure the stability of the system in (1).
First of all, the speed of adjustment parameters in one specific regime should be of opposite sign
reflecting that markets return to their equilibrium path in the long-run. From (1) it follows that
both markets can be treated as dependent simultaneously such that in each regime ∆𝑃1,𝑡 =𝜌𝑘1𝛾
′𝑃𝑡−1 and ∆𝑃2,𝑡 = 𝜌𝑘2𝛾′𝑃𝑡−1. Convergence is achieved if 𝜌𝑘1 ≤ 0 and 𝜌𝑘2 ≥ 0. Given this
restriction, it is considered sufficient that at least one adjustment parameter in a specific regime is
found significant. Secondly, the difference between the two speed of adjustment prameters of the
outer regimes should fall in the following interval 0 < 𝜌𝑘2 − 𝜌𝑘1 < 1. The last restriction
corresponds to price fluctuations decaying gradually (Greb et al., 2013).
8
We employ novel regularized Bayesian technique to identify estimates of threshold parameters,
which govern the regime switch, and restricted maximum likelihood method to estimate model
variable coefficients (Greb et al., 2013).
The presented model is estimated by two methods and within three steps. First, the long-run price
equilibrium in equation (1) is estimated by ordinary least squares (OLS) method. We retrieve the
error term which enters the TVECM lagged by one period as the ECT variable. Second, the
threshold parameters in equation (2) are identified by using the regularized Bayesian technique.
Third, the short-run and long-run price transmission parameters of equation (2) are estimated by
implementing restricted maximum likelihood method.
Compared to maximum likelihood method that utilizes maximization, the selection of thresholds
on the basis of RB estimator is done using integral calculus. According to Greb et al. (2014),
integration might be more natural to use in TVECM as it provides a means to account for inherent
variability of the estimates. A function to choose optimal threshold values over the grid of ECTs
is called posterior median and constructed as follows:
∫ 𝑃𝑅𝐵(𝜏𝑖|Δ𝑃, 𝑋)𝑑𝜏𝑖 = 0.5�̂�𝑖𝑅𝐵min (𝛾′𝑃𝑡)
, 𝑓𝑜𝑟 𝑖 = 1,2 (2‘)
where 𝑋 is a 𝑛 × 𝑑 matrix that compactly stacks together columns of ECTs and values of lagged
terms. 𝑃𝑅𝐵(𝜏|Δ𝑃, 𝑋) is well defined across the space of all possible threshold parameters Τ =
{τ1,τ2|min (𝛾′𝑃𝑡) < 𝜏1 < 𝜏2 < max (𝛾
′𝑃𝑡)}. In the previous expression, τ1 and τ2 are optimal
thresholds that separate the space into three regimes and satisfy τ1 < 0 < τ2. Computation is
based on a prior 𝑃𝑅𝐵(𝜏|𝑋) ∝ 𝐼(𝜏 ∈ 𝑇) for 𝜏, where 𝐼(∙) is an indicator function providing
switching between regimes.
Upon identification of the optimal thresholds, the additional parameters of the TVECM are
estimated. We use the restricted maximum likelihood framework implemented as a part of mixed-
effects modeling in R. Each regime is estimated independently, given the values of thresholds
(Gałecki and Burzykowski, 2013).
4.2 Identification of the determinants of market integration
Having completed price transmission analysis, next we combine price transmission elasticities
with various market characteristics in reduced-form regression analysis to identify causes of the
differences in the degree of market integration. We posit that distance and orientation of wheat
production region on export have a significant impact on the degree of market integration.
We conduct econometric analysis using Tobit model, which is fitted to the sample containing data
on regional market pairs in Russia and the USA. Model is given in the following reduced-from
equation:
𝜳𝑖 = ϑ0 + ϑ2𝓓𝑖 + ϑ3𝓧𝑖 + λ0𝓡𝑖 + 𝜆2𝓓𝑖𝓡𝑖 + 𝜆3𝓧𝑖𝓡𝑖 + 𝝃𝑖 (3)
Where 𝜳𝑖 is an estimate of the long-run price transmission elasticity from cointegration equation
(1) in Russia and the USA. 𝓓𝑖 measures average distance in kilometers between different
economic regions in Russia and between states in the USA. 𝓧𝑖 is an indicator variable and takes
value 1 if a region is an exporter to the world market, and equals to 0 otherwise. 𝓡𝑖 is a dummy
variable that equals to 1 if a observation refers to the markets in Russia and 0 – if in the USA. By
introducing interaction terms with country dummy variable 𝓡𝑖, we test conditional hypothesis
that market characteristics have different effect on market integration in Russia compared to the
USA.
5 Data sets and data properties
To estimate our price transmission model, we use a unique dataset of weekly prices of wheat of
class three (Ruble/ton), the most widely traded type of wheat for human consumption in the
Russian domestic market. This data is collected by the Russian Grain Union and is not publicly
available. The quoted prices are paid by traders to farmers on the basis of ex-works contracts. Our
9
data set comprises regional data for the six economic grain producing regions North Caucasus,
Black Earth, Central, Volga, Urals and West Siberia and contains 468 observations (January 2005
until December 2013) (Figure 3). From this database, we construct 15 market pairs in total by
combining each market with all other five regional markets in Russia.
Figure 3: Development of regional wheat prices in Russia in 2005-2013
Note: The area with dashed line on the graph covers the period of export tax (Nov 2007 - May 2008) and export ban
(Aug 2010 - Jul 2011).
Source: Own illustration, data: Russian Grain Union (2014), GTIS (2013).
In addition, we use weekly amounts of grains transported by train between all grain producing
regions of Russia as a measure for interregional grain trade flows (source: Rosstat 2015). This
data is used as additional information to build the model framework and to interpret results. As an
example, Figure 4 gives the price relationship between the regions North Caucasus and Volga as
well as North Caucasus and West Siberia (2007-2013) and the corresponding interregional grain
trade flows transported by train.
However, Figure 4 makes evident that the regional price relationships are not stable, but rather
differ from marketing year to marketing year. In particular, the price of North Caucasus is in some
period higher and in other periods lower than in the other regions. Also, the interregional trade
flows are highly volatile. This implies that the interregional price relationships, which are
depicted in the price transmission model, are highly unstable, and thus parameter estimates may
also not be constant. To tackle this issue, we estimate the price transmission model based on one
marketing year only which is characterized by relatively stable price relationships.
Figure 4: Regional wheat price relationships and interregional trade flows
0
500
1000
1500
2000
2500
3000
3500
0
2000
4000
6000
8000
10000
12000
Mar
-05
Sep
-05
Mar
-06
Sep
-06
Mar
-07
Sep
-07
Mar
-08
Sep
-08
Mar
-09
Sep
-09
Mar
-10
Sep
-10
Mar
-11
Sep
-11
Mar
-12
Sep
-12
Mar
-13
Sep
-13
10
00
t
RU
B/t
Wheat export Central Black Earth North CaucasusVolga Urals West Siberia
10
Sources: Own illustration, data: Russian Grain Union (2014).
In particular, to assess strength of market integration in Russia, we use regional price observations
of the marketing year 2009/10, when trade was freely possible, as our data base. Also, to
investigate the impact of the drought and the export ban, we estimate the price transmission model
based on the price data for the marketing year 2010/11 and compare the parameter estimates with
those obtained based on the 2009/10 price data. Both data sets comprise 52 observations each.
We use estimate long-run price transmission coefficients from free-trade regime in 2009/10 as a
dependent variable to identify determinants of market integration on the next stage. In addition,
we employ equivalent state-level corn prices for 16 states observed between marketing years 2008
and 2011 (source: USDA AMS, 2016) to get an estimate of long-run price transmission
coefficients for markets in the USA. Each price series contain 156 observations on the weekly
basis. Overall, this dataset generates 63 market pairs, which we construct by pairing 7 markets
from the major producing ‘Corn Belt’ area states with the other 9 markets mostly from net-
consumer states.
In order to identify determinants of the market integration, we supplement our dataset with the
weekly amounts of grains transported by train between all grain producing regions of Russia as a
measure for interregional grain trade flows (source: Rosstat, 2015).
Except for identifying how various market characteristics affect market functioning in Russia and
the USA, we also conduct comparative price transmission analysis to evaluate how markets
function in Russia compared to the USA, which serves as a benchmarch for the most efficient
grain market in our analysis.
Comparative price transmission analysis is conducted based on within-regional price
relationships. We select North Caucasus and West Siberia as a unit of analysis in Russia (source:
Ministry of Agriculture of Russia, 2016) and compare it with the USA markets in Iowa and North
Carolina. Figure 5 shows price developments in local markets of North Caucasus and West
Siberia.
Figure 5: Development of selected regional wheat prices in North Caucasus and West
Siberia in Russia (a), and Iowa and North Carolina, in the USA (b)
a) Russia
0
20
40
60
0
5000
10000
15000
Jan
-07
Sep
-07
May
-08
Jan
-09
Sep
-09
May
-10
Jan
-11
Sep
-11
May
-12
Jan
-13
Sep
-13
10
00
t
RU
B/t
North Caucasus & West Siberia
Exports West Siberia to North Cauc.
Price North Caucasus
Price West Siberia
-40
-20
0
20
40
0
2000
4000
6000
8000
10000
12000
Jan
-07
Sep
-07
May
-08
Jan
-09
Sep
-09
May
-10
Jan
-11
Sep
-11
May
-12
Jan
-13
Sep
-13
10
00
t
RU
B/t
North Caucasus & Volga
Exports North Cauc. to VolgaExports Volga to North Cauc.Price North CaucasusPrice Volga
11
b)
Note: Prices are by-weekly in Russia and daily in the USA. For graphical representation we select only four market in
each region.
Sources: Own illustration, data: Ministry of Agriculture of Russia (2016), GeoGrain and Nick Piggott, 2016. Own
illustration.
North Caucasus is production region which has good access to the ports and is very active on the
world market, while West Siberia – another wheat production region – is mainly active on
domestic wheat trade due to its large distances and geopraphical separation from the world
markets (Figure 6).
Figure 6. Map of crop-growing regions in North Caucasus and West Siberia
0
3500
7000
10500
14000
7/1
/20
11
1/1
/20
12
7/1
/20
12
1/1
/20
13
7/1
/20
13
1/1
/20
14
7/1
/20
14
1/1
/20
15
7/1
/20
15
1/1
/20
16
7/1
/20
16
RU
B/t
North Caucasus
Agygeya KrasnodarRostov Stavropol
0
3500
7000
10500
14000
7/1
/20
11
1/1
/20
12
7/1
/20
12
1/1
/20
13
7/1
/20
13
1/1
/20
14
7/1
/20
14
1/1
/20
15
7/1
/20
15
1/1
/20
16
7/1
/20
16
RU
B/t
West Siberia
Altai Kemerovo
Novosibirsk Omsk
0
100
200
300
400
7/1
/20
10
9/1
/20
10
11
/1/2
01
0
1/1
/20
11
3/1
/20
11
5/1
/20
11
7/1
/20
11
9/1
/20
11
11
/1/2
01
1
1/1
/20
12
3/1
/20
12
5/1
/20
12
USD
/t
Iowa
Clinton Eddyville
Emmetsburg W. Burlington
0
100
200
300
400
7/1
/20
10
9/1
/20
10
11
/1/2
01
0
1/1
/20
11
3/1
/20
11
5/1
/20
11
7/1
/20
11
9/1
/20
11
11
/1/2
01
1
1/1
/20
12
3/1
/20
12
5/1
/20
12
USD
/t
North Carolina
Candor Cofield
Creswell Laurinburg
12
Source: Own illustration.
We include price series from four winter wheat growing areas in the North Caucasus region such
as Agygeya, Krasnodar, Rostov and Stavropol. In general, price data for local markets in this
region is available from January, 2010 to September, 2016 and consists of 161 by-weekly
observations in total. Exception is Agygeya with its price series starting from June, 2013 and
accounting for 79 observations altogether.
West Siberia, which is the leading spring wheat production region in Russia and advanced grain
milling facilities, is represented by six local markets in this study. Out of them, Novosibirsk,
Altai, Omsk and Tyumen are categorized as production regions and Tomsk and Kemerovo are
net-consuming regions. Price series for Altai, Tomsk and Tyumen markets are fully available
from January, 2010 to September, 2016 including 161 by-weekly observations each. However,
other price series are given for shorter time period. More specifically, Omsk price series start in
December, 2010 generating 139 observations for this market. Price series for Novosibirsk market
starts in July, 2011 (125 observations) and for Kemerovo in September, 2012 (97 observations).
Comparable set of the data is constructed for the USA based on the prices observed in Iowa and
North Carolina (Figure 5) (source: GeoGrain and Nick Piggott, 2016). We use daily prices of the
same time period for all price series from July, 2010 to June, 2012 (506 observations each)
collected in eight markets in Iowa (Cedar Rapids, Clinton, Davenport, Eddyvile, Emmetsburg,
Keokuk, Muscatine and West Burlington) and six markets in North Carolina (Candor, Cofield,
Cresswell, Laurinburg, Roaring River and Statesville).
Before we begin with the price transmission analysis we test the properties of our price series.
Results of the Augmented Dickey-Fuller (ADF) test (Dickey and Fuller, 1981) for a unit root
(Appendix, Table A1) suggest that the all price series used in this study are integrated of order 1.
Further, we apply two different testing techniques to explore the potential of non-linearity in the
market price pairs and to confirm the use of TVECM. Hansen and Seo (2002) provide a test to
check the validity of linear co-integration under the null versus the presence of non-linear co-
integration in a two-regime TVECM with 1 threshold as the alternative. Larsen (2012) provides
an extension to the Hansen and Seo (2002) test by allowing for non-linear co-integration within a
three-regime TVECM with 2 thresholds under the alternative hypothesis.
13
Results of the two tests are given in Table A2 in the Appendix. Both Hansen and Seo (2002) and
Larsen (2012) test results suggest that the null hypothesis can be rejected at 10% level of
significance agains threshold cointegration almost in all cases. Exceptions are three market pairs
in Russia for the marketing year 2010/11 (out of 15), six matket pairs in Iowa (out of 27), and two
market pairs in North Carolina (out of 15). Overall, threshold cointegration is supported for all 15
wheat price pairs of Russia in 2009/10 by at least by one of threshold cointegration test, and for
all 6 price pairs in North Caucasus and 15 price pairs in West Siberia.
Therefore, since in the vast majority of cases test results of Hansen and Seo (2002) and Larsen
(2012) suggest threshold cointegration, we consider these results as strong evidence for the
existence of threshold effects. We explicitly account for threshold effects in the price transmission
analysis by choosing a 3-regime-TVECM for our analysis of price transmission between the
Russian regional wheat markets.
6 Results
6.1 The influence of the export ban 2010/11
The unusually low harvest in the key crop growing areas affected by the drought in 2010 induced
Russian government to impose an export ban on wheat on August 15. Initially, the ban was
introduced to last until December 2010, but it was subsequently prolonged to last until July 2011.
The measure had a profound impact on regional wheat trade in Russia. In particular, North
Caucasus could no longer supply to the world market and was forced to supply wheat
domestically instead. Table 3 shows that North Caucasus directed its flows to the markets which
suffered the most from the harvest failure, specifically Black Earth, Central, Volga and Urals.
This explains the observed wheat trade reversal, e.g. between North Caucasus and Volga region.
West Siberia was less affected by the drought and also supplied wheat to the domestic grain
producing regions which turned into deficit regions in 2010/11, in particular Volga and Urals.
Table 3. Interregional grain trade quantities by train, 2010/11
from
to
North
Caucasus
West
Siberia
Black
Earth Central Volga Urals
Regional trade (in t)
North Caucasus -2,494,506
534,336 1,205,324 453,936 300,910
West Siberia -1,180,827 73,107 101,444 1,006,276
Total imports 534,336 1,278,431 555,380 1,307,186 Source: Götz et al. 2016
To foster interregional grain trade during the export ban, the Russian government introduced
transport subsidy for grain producers located in North Caucasus starting from September 20,
2010. For example, Russian Railways cut delivery fees by half for dispatches heading from North
Caucasus towards the regions of Volga, North West and Central. The given subsidy was valid for
all grain supplies exceeding 300 kilometers and was removed together with the export ban in July
2011.
Even though railway tariff rates halfened during the export ban 2010/11, availability of trucks for
grain transportation was limited as railways were heavily involved in the construction of sport
facilities for the winter Olympic games in Sochi. Moreover, the volume of grain exported by
North Caucasus to other domestic regions was extremely high and even exceeded the availability
of trucks (Gonenko, 2011).
6.1.1 Parameters of the long-run price equilibrium regression
In this section, we discuss estimation results of price transmission analysis for Russia for the
marketing year 2009/10, when trade was freely possible, and in the marketing year 2010/11, when
Russian government imposed export ban. Table 4 presents the parameter estimates of the long-run
14
price equilibrium regression. For the marketing year 2009/10 results suggest that the long-run
price transmission parameter decreases and the intercept parameter increases with increasing
distance between the regions. This corresponds with the Law of One Price according to which
markets are perfectly integrated if the intercept of the long-run price equilibrium is equal to zero
and the slope parameter is equal to one.
Table 4: Parameters of the long-run price equilibrium regression, 2009/10 and 2010/11
Price pairs Long-run price
transmission elasticities
Intercept
parameter
Dependent
variable
Independent
variable
Distance
(km) 2009/10 2010/11 % change 2009/10 2010/11
Central Black Earth 526 0.940 0.917 -2 0.519 0.733
Central Volga 801 0.698 0.824 18 2.525 1.538
Central Urals 2044 0.432 0.670 55 4.699 2.590
Central West Siberia 3346 0.358 0.589 65 5.346 3.654
North Caucasus Black Earth 870 0.333 0.573 72 5.672 3.646
North Caucasus Central 1300 0.346 0.642 86 5.557 3.037
North Caucasus Volga 1708 0.267 0.543 103 6.225 3.896
North Caucasus Urals 2682 0.156 0.443 184 7.132 4.752
North Caucasus West Siberia 3984 0.132 0.392 197 7.340 5.262
Black Earth Volga 1035 0.740 0.890 20 2.153 0.959
Black Earth Urals 2027 0.469 0.760 62 4.366 2.052
Black Earth West Siberia 3329 0.388 0.636 64 5.071 3.248
Volga Urals 1235 0.677 0.844 25 2.645 1.326
Volga West Siberia 2537 0.571 0.717 26 3.575 2.553
Urals West Siberia 1310 0.833 0.834 0 1.452 1.590
Note: All parameters are significant at a level lower than 1%.
Source: Own estimations.
In particular, long-run price transmission is strongest between the neighbouring regions Central
and Black Earth (0.940), the first of which is the major consumption centre and the second is an
important production region, and lowest between North Caucasus and West Siberia (0.132), the
two grain producing regions which are the most distant to each other. One exception is the price
pair North Caucasus-Central, which is integrated slightly stronger than the price pair North
Caucasus-Black Earth, although Central is more distant to North Caucasus than Black Earth. The
strong integration can be explained by the regions’ trade position. Central and North Caucasus are
both the largest importing regions of Russia and strongly competing for grain imports from other
regions of Russia. Though, Central region is the main grain consuming region of Russia whereas
North Caucasus is the primary grain exporting region.
Further, it becomes evident that neighboring regions are stronger integrated than regions which
are not directly adjacent to each other. In particular, besides Central-Black Earth, Central-Volga,
Black Earth-Volga, Volga-Urals and Urals-West Siberia are the regions which exhibit
significantly stronger long-run price transmission elasticity compared to non-neighboring regions.
Our results suggest that North Caucasus is the grain producing region which is the least integrated
with the other grain producing regions of Russia. North Caucasus is the only major grain
producing region with direct access to the world grain market. Thus, different to the other grain
producing regions, North Caucasus is also strongly influenced by the world market conditions
explaining its rather low integration in the Russian regional grain markets.
For the marketing year 2010/11, when several regions experienced severe droughts, and exports to
the world market were forbidden by an export ban, the slope coefficient increases and the
intercept parameter decreases compared to 2009/10 for 13 out of the 15 price pairs. The two
exceptions are the neighboring regions Central-Black Earth and Urals-West Siberia, for which the
long-run price transmission parameter (almost) remains constant. Obviously, the domestic
15
Russian grain market is characterized by stronger market integration during the export ban. This
can be explained by two factors. First, due to the export ban, the influence of the world market
conditions on domestic price formation decreases particularly in those regions, which are usually
involved in grain export to the world market. Thus, the influence of the common domestic factors
increases, particularly in the export-oriented regions which strengthens their integration in the
domestic market. This is also reflected in the increase in the long-run price transmission
parameter (in percentage), which is strongest for the price pairs involving North Caucasus, the
increase varying between about 70% and 200%. Second, due to the severe harvest shortfalls of up
to 60% in some regions in 2010/11, interregional trade flows increase strongly and are observed
from the surplus regions North Caucasus and West Siberia to the deficit regions (compare Table
3), contributing to the strengthened domestic market integration. This rise in the domestic grain
trade was fostered by the implementation of the wheat export ban.
6.1.2 Estimated parameters of the TVECM
Selected parameters of the 3-regime TVECM, which is estimated for the 15 market pairs
separately for the marketing years 2009/10 and 2010/11 are presented in Tables 5a and Table 5b.
It becomes evident that the vast majority of observations are attributed to the middle regime for
12 out of 15 regional price pairs in 2009/10. For example, for the price pair Central–Black Earth,
40 observations are assigned to the middle regime, whereas 7 observations belong to the lower
and one observation to the upper regime. This means that the error correction term between
regional market pairs is usually smaller than the absolute value of the lower and upper threshold,
providing evidence for strong market integration. In 2010/11 the number of market pairs for
which the majority of observations lays in the middle regime increases to 14 out of the 15 market
pairs. This can be interpreted as evidence of the strengthened integration of regional markets
during the export ban.
Table 5a. Results of TVECM: Russia 2009/10
Price pair Lower regime Middle regime Upper regime Total adjustment
Number of obs.
Dependent – indep. variable Rho1 Pvalue Lower
Thresh.
Rho2 Pvalue Upper
Thresh.
Rho3 Pvalue Lower Middle Upper Band of
inaction
1 Central - Black Earth -0.212 0.360 -0.021 -0.208 0.336 0.018 -0.353 0.089 0.340 0.364 0.733 0.039
Black Earth - Central 0.340 0.072 0.364 0.035 0.380 0.015 7 40 1
2 Central - Volga -0.100 0.291 -0.013 -0.207 0.337 0.003 -0.147 0.168 - - - 0.016
Volga - Central 0.121 0.264 -0.180 0.408 -0.081 0.494 17 12 19
3 Central -Urals -0.029 0.757 -0.047 -0.149 0.259 0.029 -0.173 0.030 0.310 - 0.173 0.076
Urals - Central 0.310 0.004 0.179 0.214 0.100 0.233 17 18 13
4 Central - West Siberia -0.039 0.646 -0.062 -0.102 0.311 0.021 -0.166 0.014 0.260 - 0.166 0.083
West Siberia - Central 0.260 0.041 0.082 0.574 -0.005 0.955 12 17 19
5 North Caucasus - Black Earth -0.207 0.041 -0.021 -0.207 0.041 0.020 -0.207 0.041 0.207 0.207 0.207 0.041
Black Earth - North Caucasus -0.018 0.809 -0.018 0.809 -0.018 0.809 14 16 18
6 North Caucasus - Central -0.300 0.025 -0.030 -0.216 0.088 0.020 -0.168 0.136 0.300 0.216 - 0.050
Central - North Caucasus -0.152 0.187 0.114 0.299 -0.031 0.744 7 24 16
7 North Caucasus - Volga -0.167 0.078 -0.038 -0.177 0.136 0.012 -0.153 0.060 0.167 - 0.153 0.050
Volga - North Caucasus -0.107 0.276 -0.074 0.569 -0.091 0.328 4 26 18
8 North Caucasus - Urals 0.041 0.684 -0.036 -0.029 0.820 0.024 -0.064 0.379 - - - 0.060
Urals - North Caucasus 0.176 0.132 0.154 0.284 0.081 0.360 11 21 16
9 North Caucasus - West Siberia -0.116 0.146 -0.049 -0.125 0.036 0.029 -0.125 0.036 - 0.125 0.125 0.078
West Siberia - North Caucasus -0.010 0.926 0.057 0.573 0.057 0.573 6 29 13
10 Black Earth - Volga -0.094 0.086 -0.046 -0.146 0.052 0.011 -0.094 0.086 0.094 0.146 0.094 0.057
Volga - Black Earth 0.022 0.781 -0.003 0.979 0.022 0.781 8 26 14
11 Black Earth - Urals 0.063 0.318 -0.059 0.063 0.318 0.031 0.005 0.928 0.295 0.295 0.193 0.090
Urals - Black Earth 0.295 0.000 0.295 0.000 0.193 0.016 10 28 10
12 Black Earth - West Siberia -0.007 0.898 -0.087 -0.069 0.208 0.025 -0.049 0.375 - - - 0.112
West Siberia - Black Earth 0.106 0.229 0.015 0.859 0.016 0.849 6 26 16
13 Volga - Urals -0.160 0.203 -0.058 -0.019 0.858 0.038 -0.297 0.014 0.210 0.200 0.297 0.096
Urals - Volga 0.210 0.067 0.200 0.043 0.120 0.245 8 33 7
14 Volga - West Siberia -0.141 0.274 -0.056 -0.201 0.035 0.035 -0.288 0.004 - 0.201 0.288 0.091
West Siberia - Volga 0.216 0.125 0.098 0.228 -0.026 0.763 4 38 6
15 Urals - West Siberia -0.206 0.072 -0.027 -0.186 0.183 0.012 -0.206 0.141 0.206 - - 0.039
West Siberia - Urals 0.213 0.157 0.167 0.324 0.011 0.951 11 22 15
17
Table 5b. Results of TVECM: Russia 2010/11
Price pair Lower regime Middle regime Upper regime
Total adjustment
Number of obs.
Dependent – indep. variable Rho1 Pvalue Lower
Thresh. Rho2 Pvalue
Upper
Thresh. Rho3 Pvalue Lower Middle Upper
Band of
inaction
1 Central - Black Earth 0.018 0.964 -0.022 -0.437 0.096 0.014 -0.272 0.369 0.587 0.437 - 0.036
Black Earth - Central 0.587 0.098 0.022 0.915 0.301 0.243 6 36 6
2 Central - Volga -0.690 0.005 -0.018 -0.290 0.161 0.008 -0.168 0.334 0.690 - - 0.026
Volga - Central -0.142 0.568 0.117 0.566 0.178 0.292 8 27 13
3 Central -Urals -0.457 0.000 -0.095 0.042 0.524 0.058 -0.039 0.826 0.457 - 0.304 0.153
Urals - Central -0.017 0.873 0.084 0.171 0.304 0.078 3 41 4
4 Central -West Siberia -0.329 0.007 -0.105 0.118 0.061 0.054 0.158 0.131 0.329 -0.118 0.274 0.159
West Siberia - Central 0.040 0.772 0.028 0.764 0.274 0.042 3 38 7
5 North Caucasus - Black Earth -0.244 0.054 -0.090 -0.264 0.035 0.038 -0.217 0.121 0.244 0.264 - 0.128
Black Earth - North Caucasus -0.014 0.846 -0.075 0.171 0.008 0.921 2 38 8
6 North Caucasus - Central -0.239 0.010 -0.032 -0.385 0.397 0.004 -0.242 0.009 0.129 - 0.129 0.036
Central - North Caucasus -0.110 0.094 0.308 0.154 -0.113 0.089 16 14 18
7 North Caucasus - Volga -0.308 0.049 -0.046 -0.315 0.075 0.007 -0.260 0.066 0.054 0.315 0.103 0.053
Volga - North Caucasus -0.254 0.009 0.033 0.748 -0.157 0.042 10 23 15
8 North Caucasus - Urals -0.323 0.002 -0.099 -0.323 0.002 0.085 -0.328 0.098 0.323 0.323 0.328 0.184
Urals - North Caucasus -0.036 0.365 -0.036 0.365 -0.149 0.210 4 40 4
9 North Caucasus - West Siberia -0.381 0.000 -0.053 -0.370 0.011 0.038 -0.453 0.003 0.381 0.370 0.453 0.091
West Siberia - North Caucasus -0.048 0.536 0.013 0.921 -0.134 0.335 10 29 9
10 Black Earth - Volga -0.139 0.371 -0.029 -0.139 0.404 0.008 -0.126 0.401 - - - 0.037
Volga - Black Earth 0.012 0.948 -0.056 0.766 -0.008 0.963 6 22 20
11 Black Earth - Urals -0.271 0.011 -0.103 0.020 0.780 0.076 -0.322 0.003 0.271 - 0.322 0.179
Urals - Black Earth -0.063 0.500 0.039 0.518 -0.123 0.184 2 44 2
12 Black Earth - West Siberia -0.246 0.008 -0.107 0.041 0.430 0.071 -0.063 0.657 0.246 - - 0.178
West Siberia - Black Earth -0.150 0.186 0.104 0.126 0.003 0.984 2 44 2
13 Volga - Urals -0.194 0.027 -0.107 -0.092 0.163 0.069 -0.225 0.027 0.194 - 0.225 0.176
Urals - Volga -0.018 0.812 0.015 0.791 -0.043 0.624 2 43 3
14 Volga - West Siberia -0.104 0.170 -0.105 0.041 0.529 0.046 0.105 0.439 - - 0.418 0.151
West Siberia - Volga 0.032 0.679 0.061 0.376 0.418 0.005 4 37 7
15 Urals - West Siberia 0.053 0.513 -0.061 0.039 0.619 0.029 0.039 0.619 0.318 0.300 0.300 0.090
West Siberia - Urals 0.318 0.012 0.300 0.020 0.300 0.020 3 36 9
Note: Total adjustment in one regime is calculated as the sum of the absolute value of the respective regime-specific speed of adjustment parameters of the TVECM. The band of
inaction is given as the difference between the absolute value of the upper and lower threshold.
Source: Own estimations.
Another attribute to characterize market integration is the size of the band of inaction, difference
between the absolute value of the upper and lower threshold. The average size of the band of
inaction is significantly lower in the marketing year 2009/10 amounting to 0.07 compared to the
marketing year 2010/11 amounting to 0.12. For both marketing years the band of inaction is
highest for all price pairs which include either Ural or West Siberia, two peripheral regions which
are characterized by large distance to the grain consuming and exporting regions and thus high
trade costs. Though, the band of inaction is rather low for the price pair Urals-West Siberia, which
are neighbouring regions and are characterized by strong integration.
All price relations between the given regional markets are characterized by a positive and a
negative threshold. For example, for the market pairs containing the Central region, the threshold
with a positive value refers to the trade costs of wheat supplied to the Central region, whereas the
negative threshold corresponds to trade costs of wheat originating in the Central region and
exported to the respective partner region. As it was explained in section 2, it should be pointed out
that Central is the pivotal region representing the largest wheat consuming region of Russia, while
the other regional markets (Black Earth, West Siberia, Urals and Volga) are the primary suppliers
of wheat to Central.
Estimates of the threshold parameters in 2009/10 generally confirm the influence of distance. For
example, for all price pairs which include the Central market, the absolute value of the negative
threshold increases with distance (compare Figure 1). In particular, the absolute value of the
identified negative threshold is highest for the market pair Central-West Siberia (0.062), two
markets which are the most far apart, while it is significantly lower for the price relationship
between the markets Central and Black Earth (0.021) and Central and Volga (0.013) which are
each neighbouring regions. Thus, parameter estimates indicate that it is almost three times costlier
to supply wheat from West Siberia to Central, than from Black Earth to Central. The threshold is
second highest for the market pair Central-Urals (0.047) which is in line with the actual distance
between those markets.
A similar pattern is observed for all price pairs involving the region North Caucasus. The absolute
value of the negative threshold is lowest for the neighbouring regions North Caucasus and Black
Earth (0.021) and is highest for the most distant regions North Caucasus and West Siberia (0.049).
Generally, all price pairs including Urals or West Siberia as a region are characterized by
relatively large thresholds, which can be explained by their peripheral location and the high
transaction costs involved.
The increase in the band of inaction in 2010/11 compared to 2009/10 can be explained by the
increase of the size of thresholds. Parameter estimates suggest that the size of thresholds had
increased compared to 2009/10 for the vast majority of price pairs. The lower threshold increased
for all price pairs except one. The identified upper threshold increased for 11 out of the 15 price
pairs. These results suggest that interregional trade costs increased in 2010/11 compared to
2009/10.
Information provided by the Russian Grain Union confirms these results. First, the railway
transport costs were increased by 10% by the government in 2010/11 compared to 2009/10.
Further, the destinations of interregional grain trade flows changed during the export ban and
grain trade flows were even reversed. Traders had to extend their business to other regions and
could not make use of their established business contacts. Thus, transaction costs of trade
increased strongly by increasing trade risk associated with a high level of fraud and high risk of
contract enforcement.
The influence of distance is also reflected in the size of the regime-specific speed of adjustment
parameters and the regime-specific total adjustment. Total adjustment in one regime is calculated
as the sum of the absolute value of the respective regime-specific speed of adjustment parameters
of the TVECM. In the following we focus on the parameters which are statistically significant at
least at the 10% level, and which are of the expected sign.
19
Among the 15 price pairs, the speed of adjustment parameter is highest for the neighbouring
regions Central-Black Earth amounting to 34% to 73% per week in 2009/10 in the lower and
upper regime, respectively. The size of the speed of adjustment decreases to 31% for the price
pairs Central-Urals to 26% for Central-West Siberia, reflecting the influence of distance. The
speed of adjustment parameters observed for price pairs involving North Caucasus are
significantly lower. In particular, the highest speed of adjustment parameter is observed for the
price pair North Caucasus-Central amounting to 30%. The regime-specific parameters are
significantly lower for the price pairs North Caucasus-Black Earth, North Caucasus-Urals and
North Caucasus-West Siberia, and are decreasing with increasing distance between the regions
from 21% (North Caucasus-Black Earth) to 13% (North Caucasus West Siberia).
The influence of trade costs is also reflected when comparing the regime-specific speed of
adjustment parameters and the total adjustment for each price pair. We find 8 price pairs for
2009/10 and 12 price pairs for 2010/11 out of the 15 price pairs each for which the speed of
adjustment parameters and the total adjustment is higher in at least one of the outer regimes
(lower and upper regime) compared to the middle regime. This confirms the theory underlying
threshold models applied in spatial price transmission, according to which the speed at which
deviations from the long-run price equilibrium are corrected, is higher if the price deviations
exceed the trade costs.
The regime-specific speed of adjustment parameters are increasing for at least one regime in 13
out of 15 cases in 2010/11 compared to 2009/10, confirming once again that the integration of the
regional wheat markets was strengthened during the export ban.
6.2 Comparison with the corn market in the USA
To assess how well the regional wheat markets functions in Russia, we conduct an analysis of the
integration of the corn markets in the main grain producing regions of the USA. The corn market
of the USA seems is particularly suitable for comparison since it is characterized by rather high
variation in the level of production and strong domestic trade flows to balance supply and demand
of corn. We assume that the corn market of the USA is one of the most efficient grain markets in
the world characterized by well-developed transport and storage infrastructure and high market
transparency, serving as a benchmark for assessing the efficiency of the Russian wheat market in
this study.
Corn is the primary grain produced in the USA, accounting for more than 80% of total grain
production (USDA NASS, 2016). For comparison, wheat has a 60% share in total grain
production in Russia. The majority of corn is grown in the so-called “corn belt” region ranging
over the states Iowa, Illinois, Nebraska, Minnesota, Indiana, South Dakota, Kansas, Ohio and
Missouri and accounting for about 80% of total corn production of the USA. Similar to Russia,
the size of corn production in the USA is characterized by large regional fluctuations (Table 5).
For example, corn production in Illinois varied between 65% in 2012 and 132% in 2014 of the
average corn production of the previous 3 years. Nonetheless, the volume of harvested corn is
quite stable on the national level.
Table 5: Corn production developments in the states of the “corn-belt” area of the USA
(2004-2015), in % of the average of previous 3 years
State 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Iowa 122 106 98 110 100 111 92 104 81 101 111 118
Illinois 126 95 97 122 110 99 90 95 65 122 132 105
Nebraska 124 113 95 117 107 117 99 104 85 113 108 113
Minnesota 119 114 101 101 103 109 108 97 110 100 91 111
Indiana 121 114 97 111 97 104 97 93 67 133 132 91
South Dakota 147 111 65 123 133 150 93 105 83 137 119 113
Kansas 133 137 86 123 111 134 109 81 70 107 127 120
Ohio 126 114 98 114 86 114 106 102 85 131 114 88
Wisconsin 97 116 104 112 93 109 117 116 82 93 107 111
20
Missouri 150 94 99 118 100 111 86 88 64 135 183 100
USA total 124 108 96 117 105 111 98 98 85 117 115 105
Note: USA total considers corn production in all states.
Source: Own illustration, data: USDA NASS, 2017.
Moreover, USA is the world’s largest exporter of corn occupying 35%-40% of total corn exports
world-wide (USDA WASDE, 2017). Nonetheless, 80-90% of the corn production in the USA is
supplied to the domestic market (USDA NASS, 2017). due to the rapid expansion of biofuel
production, industrial use of the domestically produced corn has substantially increased. Even
more, since 2010 amount of corn consumed by energy sector is exceeding the quantity of corn
that is used as an animal feed. As of 2014, biofuel plants account for the 40% of total domestic
corn usage (USDA ERS, 2017). Another characteristic of the USA grain market that structurally
distinguishes it from Russia is that corn processing facilities are concentrated in the production
areas. In order to ensure logistical efficiency ethanol plants are usually established close to the
grain elevators, from where corn is primarily transported by trucks to the processing facilities.
Corn transportation in the USA is based on trucks, rails and barges (Sparger and Marathon, 2015).
On average 80% of domestic corn transfers is performed by trucks since it is the most cost
advantageous means of transportation on shorter distances (less than 500 kilometers). The rest of
the domestic corn hauling on longer distances is conducted by rails. Corn for export is mainly
transported by waterway transport. Barges are primarily busy for deliveries to the port export
terminals. They transport around 70% of totally exported corn from the “corn belt” southwards to
the Mississippi Gulf ports (FAPRI-UMC 2004), which by itself is one of the leading port for corn
export accounting for 65% of total corn exports (USDA Federal Grain Inspection Service grains
inspections, 2013, as cited in Denicoff et al., 2014). The remaining 30%-40% of the exported corn
is hauled by rail predomionantly to the Pacific Northern harbors in the Washington state.
Figure 9 illustrates cost advantages of selected transportation modes in the USA depending on the
distance covered and compares it with the railway tariff rates in Russia. Trucks are the cheapest
transportation mode in the USA within distances less than 500 kilometers. While comparing long-
distance transportation modes, barges are twice as cheap as trains if distance exceeds 1500 km,
however access to this water transport emtirely depends on the geographic proximity of a corn
trade facilities and its access to a river.
Figure 9 also shows railway tariff rates of wheat transportation in Russia to compare it with the
rates in the USA. Railway tariff rates to transport 1 tonnes of wheat in Russia are almost
undifferenciable from the rates given for corn transportation in the USA on comparable distances.
However, critical dependence on the railway infrastructure largely differs between countries.
While grain transportation over large distances solely depends on the only stated-operated railway
company in Russia, low-cost barges and several railroad companies provide corn cargo
transportation in the USA.
Figure 9. Estimated fees of grain deliveries in the USA and Russia by different modes of
transportation
0
15
30
45
60
75
0 500 1000 1500 2000 2500 3000 3500 4000
Gra
in d
eliv
ery
fees
(U
SD/t
)
Distance (km)
Comparision of unit transportation costs: Russia and the USA
Truck (USA)
Barge (USA)
Rail (USA)
Rail (Russia)
21
Notes: Barge rates represent spot shipping costs towards the southbound direction along the Mississippi River to the
ports of New Orleans from 7 different origin locations (Twin Cities, minnessota; Mid-Mississippi; Lower Illinois
River; St. Louis; Cincinnati; Lower Ohio; Cairo-Memphis) on October 05, 2010 when trade capacity and
correspondingly, railway rates are at their peak. Distance for barge rates is calculated based on National Water
Information System (2016). Rail tariff is estimated on October 01, 2010 corresponding to the peak transportation
season. Distance for Rail rates is calculated based on BNSF (2016) railway distance calculator. Truck rates are
estimated for the 3th
quarter in 2010 based on three different levels of truck rates that depend on the length of
distance: 4.15 USD per mile if distance is at most 25 miles, 2.4 USD per mile if distance is at most 100 miles, and
2.28 USD per mile if covered distance is 200 miles. Rates are based on trucks with 80 000 lbs gross vehicle weight
limit. Estiamted volume per truck is 25 metric tones.
Sources: Own illustration, data: Rosstat 2015 and USDA AMS, 2017.
6.2.1 Parameters of long-run price equilibrium relationship and TVECM estimates
Compared to Russia, transportation logistics function more efficiently and delivery costs are much
lower in USA. The influence of trade costs are usually reflected in the strength of market
integration between spatially separated regions which decrease with increasing distance between
the markets.
Table A3 and A4 (a, b, c and d) in the Appendix respectively shows estimates for the long-run
price transmission elasticities and TVECM parameter estimates for markets in North Caucasus
and West Siberia in Russia, and Iowa and North Carolina in the USA. However, in order to better
visualize comparisions across countries, as well as cross-regional differences within the countries,
we use estimation output from the Tables A3 and A4 and depict all comparisons in boxplots on
Figure 10.
Figure 10: Boxplot comparisons of price transmission coefficients for Russia and the USA
a) Long-run price transmission elasticity b) Speed of adjustment parameter
c) Observations in middle regime (%) d) Band of inaction
e) Upper thresholds f) Lower thresholds
Source: Own illustration.
Analysis shows that overall, markets in Russia and the USA function differently. Long-run price
transmission coefficients are much higher in the USA indicating almost complete transmission of
price shocks compared to Russia (panel a, Figure 10). Median estimate of price transmission
elasticities is 0.81 and 0.84 in North Caucasus and West Siberia, and 0.97 and 0.93 in Iowa and
North Carolina, respectively. In addition, similar to the results of Russian market integration
analysis on the regional level (compare table 4), price transmission coefficients are again more
heterogeneous ranging between 0.53 and 0.92 in Russia, while it has modest variation in the USA
varying between 0.81 and 1.01.
In addition, our results suggest that there are not significant differences in the functioning of grain
markets between West Siberia and North Caucasus. This is surprising as we initially expected to
observe notably lower degree of price transmission within West Siberia since the region is
separated from the world market by large distances. On the contrary, grain in North Caucasus is
only traded to export markets and local farmers have many alternatives to sell their harvest to the
export-oriented traders.
Yet being segregated from the world market, West Siberia is active in domestic wheat trade. From
the The region produces superior quality wheat and also has plentiful elevator facilities,
explaining why markets in West Siberia are functioning equally as well as in North Caucasus. The
23
only market in West Siberia which has low price transmission coefficient with the other markets
is Kemerovo province (Table A3, b in Appendix), which is net-consumer market and is less active
in the regional grain trade compared to other markets in West Siberia.
Further, efficiency of markets in transferring price shocks between regions is reflected in high
speed to correct disequilibrium. As results indicate, eliminating short-run price disequilibrium is
more time-consuming process in Russia compared to the USA (Panel b, Figure 10). In terms of
median values, price disequilibrium between market pairs is almost completely eliminated in the
USA in two weeks (93% Iowa and 89% in North Carolina), while just quarter to third of the
amount adjusted in the USA is corrected in North Caucasus (22%) and West Siberia (32%) during
the same time period.
For comparison, all market pairs in North Caucasus are within the distance interval up to 350
kilometers and show significant speed of adjustment towards equilibrium whereas distance
diapason is larger in West Siberia (from 200 to 1550 km) and market pair Kemerovo-Tyumen
with the largest distance in West Siberia does not show any adjustment at all towards long-run
equilibrium.
Another measure of the market integration is the frequency of instances when error correction
term between market pairs is smaller than the absolute value of the lower and upper threshold,
providing evidence for strong market integration. This is measured by the percent (or,
alternatively, by the number) of the ECT observations that fall in the middle regime (panel c,
Figure 10). Agreeing with the findings of the long-run price transmission analysis, Russian
markets again show lower degree of market integration compared to the USA as the median
percent of observations attributed to the middle regime is respectively 71% and 77% in North
Caucasus and West Siberia, whereas 88% and 82% of ECT values fall in the “band of inaction”
interval in case of Iowa and North Carolina.
Comparison of the band of inaction (difference between the absolute value of the upper and lower
threshold), upper and lower thresholds estimates univocally indicate that trade costs are much
larger in Russia than in the USA (panel d, e and f, Figure 10). Median threshold values vary
between 0.02 and 0.05 Iowa and North Carolina, whereas thresholds are 3-5 times larger in Russia
ranging between 0.10 and 0.14. Furthermore, threshold estiamtions show that trade costs are more
uniformly distributed in the USA and is characterized by higher variability in Russia. Difference
between 75th
and 25th
quartile values which corresponds to the height of shaded area on the
boxplots varies between 0.01 and 0.04 in the USA irrespective of the selected measure of
thresholds (upper, lower or band of inaction values). In contrast, this difference is 0.11, 0.05 and
0.11 in North Caucasus and 0.08, 0.10 and 0.19 in West Siberia for lower thresholds, upper
threshold and band of inaction, respectively. As industry practitioners indicate the key obstacles in
Russia are large distances between markets and logistical challenges that hinder intensive trade
linkages between markets of Russia.
6.2.2 Determinants of market integration and their impact in Russia and the USA
We utilize estimates of the long-run price transmission parameters from the price transmission
analysis for the regonal markets in Russia (Table 4, 2009/10) and the USA (Table A5, Appendix)
to associate it with various market characteristics and identify and compare determinants of
market integration in those countries.
Results of a formal analysis of market characteristics and are summarized in Table 6. The analysis
shows that in Russia distance has negative and statistically significant influence on price
transmission in Russia. Closer markets are more strongly integrated in Russia than markets that
are far away from each other. For instance, if we consider capital city Moscow in the Central
region as a point of reference and compare two markets in terms of proximity to Moscow, then
the one which is located 1000 km closer to the capital city will show greater magnitude of price
transmission by 18.4% than another market which is more distant from Moscow. The impact of
24
distance is less pronounced in the USA. Increase of distance between markets by 1000 km
translates into decreased price transmission coefficient only by 8.6%.
Table 6: Tobit regression results: analysis of the determinants of market integration1
Dependent variable: Parameter estimates
Long-run price transmission elasticity Russia a)
USA b)
Distance 1000 km -0.184*** -0.086***
[0.017] [0.008]
Exporter -0.353*** 0.081***
[0.055] [0.014]
Constant 0.987**
[0.019]
Observations 78
F-test (8, 70) 65.68***
(Prob > F = 0.000)
Note: a)
data sample refers to 2009/10 marketing year when trade was freely possible and includes data on 15 regional
market pairs. b)
data sample refers to 2008/11 marketing year and includes 63 observations.
*, **, *** indicate statistical significance at 10, 5 and 1%, respectively. Robust standard errors in square brackets.
Source: Own estimations.
Further, results indicate that the wheat export region of Russia is only loosely integrated with
other production regions. This is in contrast to the USA where grain prices in the export-oriented
regions strongly influences price discovery in other domestic markets in the USA. North
Caucasus, which accounts for the lion’s share of total Russian wheat export, demonstrates very
low level of market integration (on average by 35.3%) compared to other regions in Russia.
Contrary, if a region exports to the world markets in the USA, this strengthens integration of that
region with the other domestic markets by 8.1%. We interpret this result as an indicator that in the
USA market participants having easier access to the world markets consider price in exporting
region as an opportunity cost and use this information as a reference price to negotiate their own
trade transactions.
7 Conclusions
In this paper we have investigated the regional price relationships between the primary grain
production regions of Russia to assess the efficiency of the Russian wheat market and have
compared them to results for the corn market of the USA.
In general, the results of the price transmission analysis for Russia demonstrate the strong
influence of distance between the grain producing regions on their price relationships. In
particular, the band of inaction and the upper and the lower threshold increase with distance
between the regions of the price pairs, whereas the long-run price transmission elasticity, the
speed of adjustment parameter and the total adjustment decrease with distance. The speed of
adjustment parameters and total adjustment are highest for neighbouring regions.
The results for Russia also indicate high level of variation in the strength of market integration
across regions. Price pairs involving North Caucasus, the exporting region with direct access to
the world markets, are characterized by particularly low long-run price transmission elasticity,
speed of adjustment parameters and total adjustment, demonstrating that the influence of the
world market price is strongest in the exporting region North Caucasus, which reduces its regional
integration in the Russian wheat market. This suggests that the Russian grain market can be
divided in two clusters: the exporting region next to the Black Sea which is strongly influenced by
1 Due to more convenient illustration, we present complete effect of each variable separately for Russia and the USA
in this table instead of showing main effects of the variables for the USA and their incremental effect for Russia
(coefficients on country dummy interaction terms).
25
world market conditions and the other grain production regions which are almost isolated from
the exporting region and also the world market which are mainly influenced by domestic market
conditions.
In a large country like Russia, distance between the grain producing regions has strong influence
on their price relationships. The thresholds are highest for price pairs involving Urals and West
Siberia, reflecting the relatively high trade costs due to the peripheral location of those regions
within the Russian wheat market. This is reflected in a band of inaction and an upper and lower
threshold increasing with distance between the regions, whereas the long-run price transmission
elasticity, the speed of adjustment parameter and the total adjustment decrease with distance.
Thus, we find the highest speed of adjustment parameters and total adjustment for neighbouring
regions.
Our results suggest that the integration of the regional wheat markets strengthened during the
wheat export ban in 2010/11, which can be explained by the increase in interregional trade flows.
In particular, price transmission elasticities and regime-specific speed of adjustment parameters
increased in 2010/11 compared to 2009/10 for many price-pairs. Further, we find that the size of
thresholds and the band of inaction increasing in 2010/11 compared to 2009/10. We trace this
back to increasing transport costs and also increasing trade risk of interregional grain transactions.
The increasing trade risks results from the change in export destinations requiring to involve new
trade partners. Obviously, the transport subsidy was too low to prevent that total transaction costs
of interregional trade increased during the export ban period. These results confirm that in general
the risk of business is particularly high in Russia due to a high degree of fraud and the difficulties
to enforce contracts.
The comparison of the Russian wheat market with the corn market of the USA makes evident that
the efficiency of the Russian wheat market is significantly lower. In particular, the Russian market
is characterized by a high heterogeneity in the degree of price transmission compared to the USA.
Furthermore, TVECM estimations show that thresholds are larger and it takes more time for price
shocks to be corrected in Russia compared to the USA.
Our analysis on the determinants of market integration confirms a lower influence of distance on
market integration in the USA compared to Russia. Further, the exporting region is particularly
strongly integrated with other regions in the USA whereas the integration of the exporting region
in the other domestic markets is particularly low in Russia.
Our study offers several important implications in terms of trade policy and food security. First,
strengthening market integration between the grain production regions could contribute to
decrease price volatility within the regions of Russia. If price signals were faster transmitted from
deficit to surplus regions, and the transaction costs of trade were decreased, incentives for
interregional trade from surplus to the actual deficit regions would be strengthened and contribute
to cushion the price increasing effects of regional production shortfalls.
This in turn would reduce the incentives for the government to implement export controls on grain
market which in the long-run strongly negatively affect the further development of the grain
sector.
Second, the grain export potential in Russia can be increased as long as this results from an
increase in grain production in the exporting region which is well integrated in the world market.
However, the mobilization of grain export potential in other grain production regions will require
substantial investments in grain market infrastructure to improve their integration in the export
market and thus in the world grain market, which might cause substantial additional costs.
Ultimately world market price conditions will determine if this is efficient. As an alternative the
wheat supply chain might be restructured in those regions. Livestock production might settle in
the more remote grain production regions and instead of grains meat and meat products will be
exported to the world market.
26
References
BALCOMBE, K., BAILEY, A. and BROOKS, J. (2007). Threshold Effects in Price Transmission: the Case of
Brazilian Wheat, Maize and Soya Prices. American Journal of Agricultural Economics 89: 308-323.
BAYLIS, K., C. JOLEJOLE-FOREMAN and M. MALLORY (2014): Impact of wheat and rice export ban on
Indian market integration. Contributed Paper, IAMO-Forum 2014 “The rise of the “emerging
economies”: Towards functioning agricultural markets and trade relations?”, June 25-27, Halle
(Saale), Germany.
BNSF (2016): Rail Miles Inquiry, http://www.bnsf.com/
CHAN, K. S. (1993). Consistency and Limiting Distribution of the Least Squares Estimator of a Threshold
Autoregressive Model. The Annals of Statistics, 21: 520-533.
DENICOFF, M. R., PRATER, M. E., and P., BAHIZI (2014): Corn Transportation Profile, USDA Agricultural
Marketing Service, August 2014, http://dx.doi.org/10.9752/TS090.08-2014
DICKEY, D. A. and W. A. FULLER (1981): Likelihood Ratio Statistics for Autoregressive Time Series with
a Unit Root. In: Econometrica 49 (4): 1057-1072.
FACKLER, P. and B. GOODWIN (2001): Spatial price analysis. In: Gardner, B. and G. Rausser (eds.):
Handbook of Agricultural Economics. Elsevier, Amsterdam: 971-1024.
FAPRI-UMC (2004): Grain Transportation and Marketing Channels, FAPRI-UMC Briefing Paper #04-04.
GAŁECKI, A. T. and T. BURZYKOWSKI (2013). Linear mixed-effects models using R: A step-by-step
approach. New York, NY: Springer.
GEOGRAIN (2016): unpublished data, http://geograin.com/
GONENKO, I. (2011). Grain wagons versus auto. Agroinvestor, 10 (45).
GOODWIN, B. K. and N. PIGGOTT (2001): Spatial market integration in the presence of threshold effects.
In: American Journal of Agricultural Economics 83 (2): 302-317.
GÖTZ, L., T. GLAUBEN and B. BRÜMMER (2013): Wheat export restrictions and domestic market effects in
Russia and Ukraine during the food crisis. In: Food Policy 38: 214-226.
GÖTZ, L., I. DJURIC and O. NIVIEVSKYI (2016): Regional price effects of extreme weather events and
wheat export controls in Russia and Ukraine. In: Journal of Agricultural Economics 67 (3): 741-763.
GREB, F., KRIVOBOKOVA, T., MUNK, A. and S. VON CRAMON-TAUBADEL (2014). Regularized Bayesian
Estimation of Generalized Threshold Regression Models. Bayesian Anal. 9, No. 1, 171-196.
GREB, F., S. VON CRAMON-TAUBADEL, T. KRIVOBOKOVA and A. MUNK (2013): The estimation of
threshold models in price transmission analysis. In: American Journal of Agricultural Economics 95
(4): 900-916.
GTIS (2013). Global Trade Atlas, Global Trade Information Services. Unpublished data.
HANSEN, B. and B. SEO (2002): Testing for two-regime threshold co-integration in vector error-correction
models. In: Journal of Econometrics 110 (2): 293-318.
INTERFAX, 2016. Russia & CIS Food and Agriculture Weekly, various issues.
JOHANSEN, S. (1988): Statistical Analysis of Cointegration Vectors. In: Journal of Economic Dynamics
and Control 12 (2-3): 231-254.
LARSEN, B. (2012): A Threshold Cointegration Analysis of Norwegian Interest Rates. Master Thesis.
University of Tromsø.
LIOUBIMTSEVA, E. and G. HENEBRY (2012): Grain production trends in Russia, Ukraine and Kazakhstan:
New opportunities in an increasingly unstable world? In: G.M. Front. Earth Sci. 6 (2): 157-166.
LO, M.C. and E. ZIVOT (2001): Threshold cointegration and nonlinear adjustment to the Law of One Price.
In: Macroeconomic Dynamics 5 (4): 533-576.
MACKINNON, J. G. (1996). Numerical Distribution Functions for Unit Root and Cointegration Tests.
Journal of Applied Econometrics, 11, 601-618.
MINISTRY OF AGRICULTURE OF RUSSIA (2016): Специализированный Центр Учета В
Агропромышленном Комплексе, http://www.specagro.ru/
NATIONAL WATER INFORMATION SYSTEM (2016): NWIS Mapper,https://maps.waterdata.usgs.gov/mapper
OECD/FAO (2012): OECD-FAO Agricultural Outlook 2012-2021. OECD Publishing and FAO.
27
ROSSTAT (2015): Federal State Statistics Service. www.gks.ru.
RUSSIAN GRAIN UNION (2014). Unpublished data.
SCHERBANIN, Y. (2012). Some problems of development of railway infrastructure in Russia (in Russian
lnguage). Problems of Forecasting, 12: 49-62.
SPARGER, A. and N., MARATHON (2015). Transportation of U.S. Grains: A Modal Share Analysis 1978-
2013 update, June 2015. USDA, Agricultural Marketing Service.
STEPHENS, E. C., MABAYA, E., VON CRAMON-TAUBADEL, S. and C. B. BARRETT (2012). Spatial price
adjustment with and without trade. Oxford Bulletin of Economics and Statistics, 74-3: 453-469.
SURFACE TRANSPORTATION BOARD (2016): Carload Waybill Samples. www.stb.gov.
USDA AMS (2016): Market News online database, https://www.marketnews.usda.gov
USDA AMS (2017): Transportation Services Division, https://www.ams.usda.gov/services/transportation-
analysis
USDA ERS (2017): Feed Grains Database, https://www.ers.usda.gov/data-products/feed-grains-database/
USDA NASS (2016): Crop Production 2015 Summary (January, 2016).
USDA NASS (2017): Quick stats database, https://quickstats.nass.usda.gov/
USDA WASDE (2017): World Agricultural Supply and Demand Estimates, WASDE-563 (March 9, 2017)
https://www.usda.gov/oce/commodity/wasde/
Appendix
Table A1. Augmented Dickey-Fuller test for prices in levels and first differences
Variable Determ.
component
Lags Test-stat Δ Variable Determ.
component
Lags Test-stat
Russia (regional)
07/2005 – 12/2012
Central Constant & trend 2 -2.531 Δ Central None 1 -7.396***
North Caucasus Constant & trend 1 -2.287 Δ North Caucasus None 0 -10.14***
Black Earth Constant & trend 1 -2.362 Δ Black Earth None 0 -8.520***
Volga Constant & trend 2 -2.569 Δ Volga None 1 -7.252***
Urals Constant & trend 2 -2.380 Δ Urals None 1 -7.351***
West Siberia Constant & trend 2 -2.546 Δ West Siberia None 1 -7.349***
North Caucasus
07/2011 – 09/2016
Adygeya None 1 -0.085 Δ Adygeya None 0 -9.628***
Krasnodar Constant 1 -2.412 Δ Krasnodar None 0 -8.827***
Rostov Constant & trend 1 -2.674 Δ Rostov None 0 -9.943***
Stavropol Constant 0 -2.022 Δ stavropol None 0 -10.41***
West Siberia
07/2011 – 09/2016
Altai Constant 1 -2.356 Δ Altai None 0 -6.979***
Kemerovo Constant 1 -2.343 Δ Kemerovo None 0 -9.098***
Novosibirsk Constant & trend 0 -1.851 Δ Novosibirsk None 0 -9.688***
Omsk Constant & trend 1 -2.712 Δ Omsk None 0 -8.317***
Tomsk Constant 1 -2.052 Δ Tomsk None 0 -6.737***
Tyumen Constant & trend 0 -2.460 Δ Tyumen None 0 -11.38***
USA (regional)
07/2008 – 06/2011
Tyumen
Arkansas Constant & trend 1 -3.018 Δ Arkansas None 0 -10.22***
California Constant & trend 0 -3.080 Δ California None 0 -11.94***
Colorado Constant & trend 1 -2.548 Δ Colorado None 0 -10.19***
Illinois Constant & trend 1 -2.816 Δ Illinois None 0 -10.14***
Iowa Constant & trend 1 -2.753 Δ Iowa None 0 -10.10***
Kansas Constant & trend 1 -2.607 Δ Kansas None 0 -9.961***
Minnesota Constant & trend 0 -2.940 Δ Minnesota None 0 -10.31***
Missouri Constant & trend 0 -3.144* Δ Missouri None 0 -10.36***
Nebraska Constant & trend 0 -2.849 Δ Nebraska None 0 -10.54***
Oklahoma Constant & trend 0 -2.878 Δ Oklahoma None 0 -10.96***
28
Oregon Constant & trend 0 -3.088 Δ Oregon None 0 -10.32***
S. Dakota Constant & trend 0 -2.822 Δ S. Dakota None 0 -10.71***
Texas Constant & trend 0 -2.948 Δ Texas None 0 -10.25***
Virginia Constant & trend 0 -3.038 Δ Virginia None 0 -10.35***
Washington Constant & trend 0 -3.152* Δ Washington None 0 -10.41***
Wyoming Constant & trend 0 -2.688 Δ Wyoming None 0 -10.50***
Iowa
07/2010 – 06/2012
Cedar Rapids Constant & trend 0 -2.915 Δ Cedar Rapids None 0 -21.64***
Clinton Constant & trend 0 -3.031 Δ Clinton None 0 -21.91***
Davenport Constant & trend 0 -3.212* Δ Davenport None 0 -21.02***
Eddyville Constant & trend 0 -2.944 Δ Eddyville None 0 -22.29***
Emmetsburg Constant & trend 0 -3.026 Δ Emmetsburg None 0 -21.91***
Keokuk Constant & trend 0 -3.266* Δ Keokuk None 0 -21.14***
Muscatine Constant & trend 0 -3.127* Δ Muscatine None 0 -21.48***
W. Burlington Constant & trend 0 -3.268* Δ W. Burlington None 0 -21.32***
North Carolina
07/2010 – 06/2012
Candor Constant & trend 0 -2.844 Δ Candor None 0 -23.36***
Cofield Constant & trend 0 -2.984 Δ Cofield None 0 -22.34***
Creswell Constant & trend 0 -2.842 Δ Creswell None 0 -23.48***
Laurinburg Constant & trend 0 -2.906 Δ Laurinburg None 0 -22.33***
Roaring River Constant & trend 1 -2.916 Δ Roaring River None 0 -27.18***
Rose Hill Constant & trend 0 -2.906 Δ Rose Hill None 0 -22.32***
Statesville Constant & trend 0 -3.123* Δ Statesville None 0 -21.88***
Note: Prices are given in natural logarithm. The asterisks refer to the significance at 1% (***), 5% (**) and 10% (*)
levels. Lag length selection is based on Schwarz Information Criterion. One-sided p-values are from MacKinnon
(1996).
Source: Own calculations.
Table A2. Tests of threshold cointegration:
a) Russia (regional) 2009/10
Price series
Specification Hansen & Seo test
(2002) a
Larsen test
(2012) b
Intercept Lags Sup-Wald
test statistic
5% critical
value P-value
Central – Black Earth yes 1 11.111 18.398 0.061
Central – Volga yes 2 20.817* 21.537 0.512
Central – Urals no 2 20.363*** 18.295 0.140
Central – West Siberia yes 2 19.219* 20.598 0.100
North Caucasus – Central no 2 21.037*** 20.011 0.033
North Caucasus – Black Earth no 1 13.932* 14.233 0.082
North Caucasus – Volga no 2 21.666*** 18.378 0.043
North Caucasus – Urals no 2 24.227*** 18.548 0.008
North Caucasus – West Siberia no 2 20.543** 19.168 0.040
Black Earth – Volga yes 3 24.383* 05.088 0.070
Black Earth – Urals yes 3 25.332*** 24.907 0.010
Black Earth – West Siberia yes 1 15.223* 16.237 0.080
Volga – Urals no 2 17.746* 18.340 0.417
Volga – West Siberia no 1 12.149* 13.192 0.076
Urals – West Siberia no 2 18.002* 18.360 0.507 Note: Sample runs from 7/3/2009 to 6/25/2010 (52 obs.)
b) Russia (regional) 2010/11
Price series
Specification Hansen & Seo test
(2002) a
Larsen test
(2012) b
Intercept Lags Sup-Wald
test statistic
5% critical
value P-value
29
Central – Black Earth no 2 18.477** 18.042 0.032
Central – Volga no 2 18.477*** 17.512 0.027
Central – Urals no 2 20.360** 19.903 0.080
Central – West Siberia yes 1 16.407* 17.643 0.358
North Caucasus – Central yes 1 15.189** 14.963 0.081
North Caucasus – Black Earth yes 1 15.038* 15.524 0.047
North Caucasus – Volga no 3 23.181** 23.167 0.030
North Caucasus – Urals no 2 23.871*** 17.998 0.130
North Caucasus – West Siberia yes 1 14.983 17.988 0.264
Black Earth – Volga no 3 23.722** 13.249 0.446
Black Earth – Urals no 1 11.489 13.571 0.169
Black Earth – West Siberia no 3 25.341*** 23.204 0.040
Volga – Urals no 3 24.684** 23.650 0.203
Volga – West Siberia no 3 24.313** 23.285 0.108
Urals – West Siberia no 1 8.536 13.635 0.132 Note: Sample runs from 7/2/2010 to 6/24/2011 (52 obs.)
d) North Caucasus (Russia)
Price series
Specification Hansen & Seo test
(2002) a
Larsen test
(2012) b
Intercept Lags Sup-Wald
test statistic
5% critical
value P-value
Adygea – Krasnodar no 5 33.365** 33.132 0.017
Adygea – Rostov yes 5 16.635 22.353 0.071
Adygea – Stavropol yes 5 29.552* 31.008 0.032
Krasnodar – Rostov no 5 19.929*** 14.370 0.005
Krasnodar – Stavropol no 5 14.829** 14.603 0.059
Rostov – Stavropol no 5 19.970*** 14.208 0.002 Note: Sample runs from 7/01/2011 to 9/01/2016 (125 obs.). Sample containing Adygea runs from 7/01/2013 to 9/01/2016 (77
obs.).
e) West Siberia (Russia)
Price series
Specification Hansen & Seo test
(2002) a
Larsen test
(2012) b
Intercept Lags Sup-Wald
test statistic
5% critical
value P-value
Kemerovo – Altai yes 2 23.137** 22.828 0.039
Kemerovo – Novosibirsk yes 2 27.360*** 21.975 0.029
Kemerovo – Omsk no 1 20.482*** 14.081 0.009
Kemerovo – Tomsk no 2 19.149* 20.155 0.067
Novosibirsk – Altai no 2 19.790* 20.568 0.307
Novosibirsk – Omsk no 1 19.122*** 14.382 0.009
Novosibirsk – Tyumen no 1 11.444 14.608 0.037
Tomsk – Novosibirsk no 1 19.151*** 15.103 0.013
Tomsk – Altai no 2 21.258** 19.785 0.126
Tomsk – Omsk no 1 19.094*** 14.804 0.004
Tomsk – Tyumen no 1 16.832** 15.465 0.044
Altai – Omsk yes 3 27.514* 27.819 0.021
Altai – Tyumen no 1 15.165* 15.171 0.003
Tyumen – Kemerovo yes 2 22.884** 21.928 0.002
Tyumen – Omsk no 1 20.173*** 15.508 0.001 Note: Sample runs from 7/01/2011 to 9/01/2016 (125 obs.). Sample with Kemerovo runs from 9/01/2012 to 9/01/2016 (97 obs.).
f) Iowa (the USA)
Price series Specification Hansen & Seo test
(2002) a
Larsen test
(2012) b
30
Intercept Lags Sup-Wald
test statistic
5% critical
value P-value
Cedar Rapids – Emmetsburg no 3 24.431 28.050 0.217
Clinton – Cedar Rapids no 2 27.187*** 22.215 0.074
Clinton – Davenport no 2 24.477** 23.376 0.066
Clinton – Emmetsburg no 2 19.137 23.281 0.061
Clinton – Muscatine yes 2 25.395* 26.041 0.209
Davenport – Cedar Rapids no 2 20.054 23.636 0.176
Davenport – S. Emmetsburg no 2 32.241*** 23.040 0.001
Eddyville – Cedar Rapids no 2 25.386** 23.167 0.019
Eddyville – Clinton no 2 21.868* 22.735 0.074
Eddyville – Davenport no 2 35.571*** 23.287 0.000
Eddyville – Emmetsburg no 2 29.665*** 23.347 0.013
Eddyville – Keokuk no 2 30.147*** 23.205 0.007
Eddyville – Muscatine no 4 28.437 33.871 0.190
Keokuk – Cedar Rapids no 4 31.472 33.946 0.250
Keokuk – Clinton no 1 25.649*** 16.836 0.001
Keokuk – Davenport no 2 30.148*** 23.290 0.006
Keokuk – Emmetsburg no 2 20.434 22.508 0.153
Keokuk – Muscatine no 2 26.417*** 22.629 0.026
Muscatine – Cedar Rapids no 2 32.400*** 17.332 0.000
Muscatine – Davenport no 2 21.769* 23.465 0.020
Muscatine – Emmetsburg no 2 30.974*** 22.924 0.016
West Burlington – Cedar Rapids no 4 39.530*** 33.104 0.015
West Burlington – Clinton no 2 20.466 22.638 0.342
West Burlington – Davenport no 4 36.641** 34.202 0.033
West Burlington – Eddyville no 4 39.114*** 33.906 0.026
West Burlington – Emmetsburg no 2 21.803* 22.424 0.290
West Burlington – Keokuk no 4 31.696* 33.709 0.059 Note: Sample runs from 6/25/2010 to 6/01/2012 (506 obs.). The asterisks refer to the significance at 1% (***), 5% (**) and 10%
(*) levels.
g) North Carolina (the USA)
Price series
Specification Hansen & Seo test
(2002) a
Larsen test
(2012) b
Intercept Lags Sup-Wald
test statistic
5% critical
value P-value
Candor – Creswell no 2 23.120** 21.675 0.007
Cofield – Candor no 1 23.179*** 17.185 0.005
Cofield – Creswell no 1 34.182*** 17.031 0.000
Laurinburg – Candor no 3 30.420** 27.055 0.150
Laurinburg – Cofield no 3 31.259** 28.029 0.134
Laurinburg – Creswell no 2 24.169** 22.374 0.064
Laurinburg – Roaring River no 3 20.899 28.304 0.604
Laurinburg – Statesville no 3 30.895** 27.819 0.016
Roaring River – Candor no 3 33.271*** 27.880 0.283
Roaring River – Cofield no 4 31.830* 33.638 0.007
Roaring River – Creswell no 4 30.025 33.727 0.533
Roaring River – Statesville no 3 28.335 28.316 0.033
Statesville – Candor no 3 35.062*** 27.807 0.060
Statesville – Cofield no 2 26.437*** 22.143 0.009
Statesville – Creswell no 2 44.873*** 21.756 0.000 Note: Sample runs from 6/25/2010 to 6/01/2012 (506 obs.). The asterisks refer to the significance at 1% (***), 5% (**) and 10%
(*) levels.a: H0: linear cointegration | H1: threshold cointegration. 1 threshold, trimming parameter is 0.05, number of bootstrapping
is set to 1000, type of bootstrapping is ‘fixed Regression’. b: H0: linear cointegration | H1: threshold cointegration. 2 thresholds, trimming parameter is 0.05, number of bootstrapping is set to
1000, type of bootstrapping is ‘fixed Regression’.
31
Source: Own estimations.
Table A3: Parameters of the long-run price equilibrium regression
a) North Caucasus (Russia)
Dependent variable Independent
variable Distance (km)
Long-run price
transmission elasticities
Intercept
parameter
Adygea Krasnodar 132 0.915*** 0.777
Adygea Rostov 344 0.863*** 1.182
Adygea Stavropol 229 0.763*** 2.159***
Krasnodar Rostov 277 0.867*** 1.165***
Krasnodar Stavropol 297 0.745*** 2.309***
Rostov Stavropol 343 0.799*** 1.863***
b) West Siberia (Russia)
Dependent variable Independent
variable Distance (km)
Long-run price
transmission elasticities
Intercept
parameter
Kemerovo Altai 411 0.526*** 4.925***
Kemerovo Novosibirsk 267 0.649*** 3.219***
Kemerovo Omsk 906 0.549*** 4.134***
Kemerovo Tomsk 218 0.595*** 3.632***
Novosibirsk Altai 226 0.916*** 0.690**
Novosibirsk Omsk 654 0.843*** 1.440***
Novosibirsk Tyumen 1280 0.888*** 1.055***
Tomsk Novosibirsk 268 0.876*** 1.230***
Tomsk Altai 490 0.920*** 0.782***
Tomsk Omsk 911 0.807*** 1.879***
Tomsk Tyumen 1538 0.841*** 1.589***
Altai Omsk 880 0.861*** 1.338***
Altai Tyumen 1504 0.882*** 1.173***
Tyumen Kemerovo 1548 0.879*** 1.016
Tyumen Omsk 624 0.832*** 1.480***
Tyumen Omsk 624 0.832*** 1.480***
c) Iowa (the USA)
Dependent variable Independent
variable Distance (km)
Long-run price
transmission elasticities
Intercept
parameter
Cedar Rapirs Emmetsburg 354 0.918*** 0.483***
Clinton Cedar Rapids 138 0.984*** 0.097***
Clinton Davenport 66 0.979*** 0.139***
Clinton Emmetsburg 489 0.905*** 0.564***
Clinton Muscatine 114 0.984*** 0.100***
Davenport Cedar Rapids 129 0.992*** 0.027
Davenport Emmetsburg 483 0.919*** 0.459***
Eddyville Cedar Rapids 174 0.964*** 0.190***
Eddyville Clinton 290 0.978*** 0.104***
Eddyville Davenport 240 0.963*** 0.210***
Eddyville Emmetsburg 367 0.889*** 0.633***
Eddyville Keokuk 182 0.991*** 0.036
Eddyville Muscatine 166 0.966*** 0.181***
Keokuk Cedar Rapids 188 0.968*** 0.180***
Keokuk Clinton 253 0.982*** 0.096***
Keokuk Davenport 190 0.969*** 0.191***
Keokuk Emmetsburg 542 0.893*** 0.626***
Keokuk Muscatine 140 0.971*** 0.170***
Muscatine Cedar Rapids 105 0.994*** 0.031
Muscatine Davenport 48 0.991*** 0.063**
32
Muscatine Emmetsburg 462 0.916*** 0.490***
West Burlington Cedar Rapids 159 0.963*** 0.203***
West Burlington Clinton 193 0.978*** 0.115***
West Burlington Davenport 126 0.965*** 0.211***
West Burlington Eddyville 151 0.997*** 0.008
West Burlington Emmetsburg 512 0.888*** 0.647***
West Burlington Keokuk 66 0.993*** 0.035**
d) North Carolina (the USA)
Dependent variable Independent
variable Distance (km)
Long-run price
transmission elasticities
Intercept
parameter
Candor Creswell 360 0.864*** 0.846***
Cofield Candor 333 0.994*** -0.009
Cofield Creswell 97 0.872*** 0.760***
Laurinburg Candor 71 0.925*** 0.396***
Laurinburg Cofield 343 0.927*** 0.426***
Laurinburg Creswell 370 0.812*** 1.110***
Laurinburg Roaring River 261 0.968*** 0.159***
Laurinburg Statesville 211 0.906*** 0.536***
Roaring River Candor 192 0.943*** 0.321***
Roaring River Cofield 286 0.935*** 0.400***
Roaring River Creswell 475 0.819*** 1.093***
Roaring River Statesville 65 0.918*** 0.488***
Statesville Candor 157 1.009*** -0.082*
Statesville Cofield 439 1.005*** -0.018
Statesville Creswell 470 0.878*** 0.736***
Note: The asterisks refer to the significance at 1% (***), 5% (**) and 10% (*) levels.
Source: Own estimations.
Table A4. Results of TVECM:
a) North Caucasus, Russia
Price pair Lower regime Middle regime Upper regime Total adjustment
Number of obs.
Dependent – indep. variable 𝝆𝟏 [Pvalue] Lower
Thresh.
𝝆𝟐 [Pvalue] Upper
Thresh.
𝝆𝟑 [Pvalue] Lower Middle Upper Band of
inaction
1 Adygeya - Krasnodar -0.306 0.000 -0.109 0.027 0.475 0.084 -0.078 0.196 0.306 - - 0.192
Krasnodar - Adygeya 0.115 0.330 -0.066 0.441 0.133 0.215 8 47 14
2 Adygeya - Stavropol -0.222 0.002 -0.131 -0.103 0.122 0.107 -0.095 0.137 0.222 - - 0.238
Stavropol - Adygeya -0.130 0.057 0.036 0.570 0.050 0.414 11 53 10
3 Adygeya -Urals -0.237 0.001 -0.136 -0.111 0.043 0.089 -0.111 0.043 0.237 0.111 0.111 0.224
Urals - Adygeya -0.036 0.679 0.131 0.142 0.131 0.142 7 52 15
4 Krasnodar - Rostov -0.513 0.064 -0.140 -0.148 0.018 0.143 0.316 0.748 0.513 0.148 - 0.283
Rostov - Krasnodar -0.689 0.032 -0.086 0.302 0.989 0.321 3 117 2
5 Krasnodar - Stavropol -0.117 0.458 -0.223 -0.151 0.016 0.122 -0.129 0.102 0.486 0.384 0.129 0.345
Stavropol - Krasnodar 0.486 0.007 0.232 0.012 0.059 0.580 1 114 8
6 Rostov - Stavropol -0.164 0.359 -0.283 -0.085 0.289 0.041 -0.130 0.158 0.730 0.219 0.148 0.325
Stavropol - Rostov 0.730 0.000 0.219 0.006 0.148 0.011 1 84 38
b) West Siberia, Russia
Price pair Lower regime Middle regime Upper regime Total adjustment
Number of obs.
Dependent – indep. variable 𝝆𝟏 [Pvalue] Lower
Thresh.
𝝆𝟐 [Pvalue] Upper
Thresh.
𝝆𝟑 [Pvalue] Lower Middle Upper Band of
inaction
1 Kemerovo – Altai -0.204 0.037 -0.176 -0.198 0.038 0.049 -0.172 0.066 0.204 0.198 0.172 0.226
Altai – Kemerovo -0.127 0.225 -0.096 0.353 -0.188 0.062 4 60 30
2 Kemerovo – Novosibirsk -0.483 0.006 -0.085 -0.063 0.576 0.060 -0.063 0.576 0.483 0.331 0.331 0.146
Novosibirsk – Kemerovo 0.151 0.429 0.331 0.014 0.331 0.014 8 73 13
3 Kemerovo – Omsk -0.318 0.026 -0.120 -0.191 0.038 0.147 -0.366 0.015 0.318 0.191 0.366 0.267
Omsk – Kemerovo -0.012 0.940 0.124 0.319 -0.175 0.317 5 86 2
4 Kemerovo – Tomsk -0.786 0.000 -0.099 -0.042 0.578 0.088 -0.491 0.018 0.786 - 0.491 0.187
Tomsk – Kemerovo -0.285 0.218 0.013 0.917 -0.577 0.020 11 78 5
5 Novosibirsk – Altai -0.456 0.000 -0.257 -0.364 0.000 0.157 -0.333 0.000 0.456 0.364 0.333 0.413
Altai – Novosibirsk -0.076 0.297 -0.047 0.459 -0.037 0.518 4 108 10
6 Novosibirsk – Omsk -0.304 0.001 -0.166 -0.149 0.009 0.192 -0.310 0.002 0.304 0.383 0.310 0.358
Omsk – Novosibirsk 0.099 0.410 0.235 0.003 0.117 0.335 7 112 4
7 Novosibirsk – Tyumen 0.126 0.321 -0.364 -0.133 0.010 0.105 -0.042 0.450 0.310 0.133 0.157 0.469
34
Tyumen – Novosibirsk 0.310 0.030 0.085 0.201 0.157 0.027 1 93 29
8 Tomsk – Novosibirsk -0.024 0.719 -0.128 0.026 0.738 0.157 0.024 0.769 0.329 0.243 0.416 0.285
Novosibirsk – Tomsk 0.329 0.000 0.243 0.001 0.416 0.000 11 107 5
9 Tomsk – Altai -0.386 0.001 -0.047 -0.497 0.002 0.014 -0.362 0.000 0.386 0.497 0.362 0.061
Altai – Tomsk 0.015 0.896 -0.084 0.606 0.092 0.374 16 61 45
10 Tomsk – Omsk -0.194 0.025 -0.068 -0.079 0.447 0.105 -0.049 0.547 0.194 0.261 0.407 0.172
Omsk – Tomsk 0.155 0.120 0.261 0.026 0.407 0.000 20 95 8
11 Tomsk – Tyumen 0.017 0.805 -0.176 -0.024 0.558 0.132 -0.002 0.971 0.361 0.214 0.224 0.308
Tyumen – Tomsk 0.361 0.000 0.214 0.001 0.224 0.003 7 94 22
12 Altai – Omsk -0.177 0.044 -0.113 0.216 0.048 0.019 0.104 0.221 0.177 0.441 0.461 0.132
Omsk – Altai 0.165 0.118 0.441 0.001 0.461 0.000 11 51 59
13 Altai – Tyumen -0.072 0.138 -0.225 0.008 0.816 0.171 0.008 0.816 0.241 0.208 0.208 0.396
Tyumen – Altai 0.241 0.001 0.208 0.001 0.208 0.001 6 100 16
14 Tyumen – Kemerovo -0.089 0.368 -0.109 -0.175 0.157 0.101 -0.089 0.368 - - - 0.210
Kemerovo – Tyumen 0.110 0.128 -0.086 0.358 0.110 0.128 19 48 17
15 Tyumen – Omsk -0.322 0.000 -0.114 -0.318 0.000 0.125 -0.322 0.000 0.322 0.318 0.322 0.239
Omsk – Tyumen -0.002 0.972 -0.004 0.951 -0.002 0.972 27 74 12
c) Iowa, the USA
Price pair Lower regime Middle regime Upper regime Total adjustment
Number of obs.
Dependent – indep. variable 𝝆𝟏 [Pvalue] Lower
Thresh.
𝝆𝟐 [Pvalue] Upper
Thresh.
𝝆𝟑 [Pvalue] Lower Middle Upper Band of
inaction
1 Cedar Rapids – Emmetsburg -0.812 0.063 -0.012 -0.807 0.067 0.008 -0.812 0.063 0.812 0.807 0.812 0.020
Emmetsburg – Cedar Rapids -0.554 0.372 -0.549 0.381 -0.554 0.372 121 226 153
2 Clinton – Cedar Rapids -0.959 0.387 -0.024 -0.996 0.053 0.012 -0.887 0.552 - 0.996 - 0.037
Cedar Rapids – Clinton 0.983 0.279 -0.695 0.621 0.982 0.303 3 481 16
3 Clinton – Davenport -0.891 0.083 -0.009 0.818 0.130 0.039 -0.891 0.083 0.891 - 0.891 0.048
Davenport – Clinton -0.204 0.851 0.239 0.803 -0.204 0.851 158 334 8
4 Clinton – Emmetsburg -0.879 0.046 -0.001 0.795 0.191 0.021 -0.921 0.028 0.879 - 0.921 0.021
Emmetsburg – Clinton -0.648 0.317 0.452 0.613 -0.702 0.280 254 190 56
5 Clinton – Muscatine 0.783 0.477 -0.025 0.825 0.293 0.017 0.629 0.618 0.999 - 0.998 0.042
Muscatine – Clinton 0.999 0.010 0.808 0.313 0.998 0.009 9 460 31
6 Davenport – Cedar Rapids 0.418 0.662 -0.032 0.491 0.523 0.028 -0.491 0.523 0.899 - - 0.059
Cedar Rapids – Davenport 0.899 0.075 0.468 0.537 0.468 0.537 26 460 14
7 Davenport – Emmetsburg -0.917 0.059 -0.023 -0.960 0.022 0.010 -0.921 0.054 0.917 0.960 0.921 0.033
Emmetsburg – Davenport -0.319 0.754 0.711 0.346 -0.372 0.706 18 377 105
8 Eddyville – Cedar Rapids 0.460 0.733 -0.045 0.626 0.373 0.028 0.460 0.733 0.986 - 0.986 0.073
35
Cedar Rapids – Eddyville 0.986 0.035 0.541 0.463 0.986 0.035 6 473 21
9 Eddyville – Clinton -0.516 0.666 -0.028 0.506 0.538 0.032 -0.114 0.945 - - 0.988 0.060
Clinton – Eddyville 0.927 0.138 0.642 0.363 0.988 0.027 20 462 18
10 Eddyville – Davenport -0.688 0.238 -0.002 0.750 0.463 0.001 -0.637 0.282 - - - 0.003
Davenport – Eddyville 0.429 0.561 0.387 0.788 0.486 0.473 252 37 211
11 Eddyville – Emmetsburg -0.958 0.016 -0.034 -0.900 0.034 0.022 -0.668 0.165 0.958 0.900 - 0.056
Emmetsburg – Eddyville -0.828 0.163 -0.689 0.267 0.100 0.891 20 400 83
12 Eddyville – Keokuk -0.981 0.026 -0.026 -0.954 0.031 0.014 -0.880 0.143 0.981 0.954 - 0.041
Keokuk – Eddyville -0.299 0.819 0.425 0.668 0.735 0.340 17 414 72
13 Eddyville – Muscatine 0.058 0.965 -0.031 0.329 0.725 0.023 0.116 0.920 - - 0.900 0.054
Muscatine – Eddyville 0.876 0.139 0.729 0.251 0.900 0.075 11 449 40
14 Keokuk – Cedar Rapids -0.603 0.574 -0.033 0.082 0.938 0.040 -0.651 0.542 - - - 0.073
Cedar Rapids – Keokuk 0.810 0.326 0.828 0.128 0.814 0.340 14 485 1
15 Keokuk – Clinton 0.999 0.090 -0.060 0.189 0.845 0.033 0.999 0.090 0.001 - 0.060 0.093
Clinton – Keokuk 0.999 0.010 0.830 0.116 0.999 0.010 1 492 7
16 Keokuk – Davenport -0.928 0.073 -0.026 0.087 0.940 0.016 -0.917 0.090 0.928 - 0.917 0.042
Davenport – Keokuk 0.279 0.815 0.877 0.121 0.333 0.773 13 425 62
17 Keokuk – Emmetsburg -0.553 0.289 -0.011 -0.553 0.289 0.035 -0.553 0.289 - - - 0.046
Emmetsburg – Keokuk 0.259 0.695 0.259 0.695 0.259 0.695 139 343 21
18 Keokuk – Muscatine 0.595 0.415 -0.021 0.589 0.411 0.027 0.592 0.416 0.924 0.917 0.922 0.048
Muscatine – Keokuk 0.924 0.027 0.917 0.028 0.922 0.028 22 455 23
19 Muscatine – Cedar Rapids -0.889 0.260 -0.041 -0.989 0.008 0.012 -0.915 0.207 - 0.989 - 0.053
Cedar Rapids – Muscatine 0.735 0.483 -0.678 0.459 0.557 0.659 2 424 74
20 Muscatine – Davenport -0.907 0.024 -0.008 -0.905 0.025 0.011 -0.905 0.025 0.907 0.905 0.905 0.019
Davenport – Muscatine -0.457 0.548 -0.449 0.559 -0.449 0.559 170 202 128
21 Muscatine – Emmetsburg -0.910 0.021 -0.013 -0.910 0.021 0.012 -0.918 0.009 0.910 0.910 0.918 0.024
Emmetsburg – Muscatine -0.729 0.204 -0.729 0.204 -0.722 0.180 109 277 114
22 W. Burlington – Cedar Rapids 0.999 0.034 -0.066 -0.537 0.470 0.039 0.999 0.034 0.999 - 0.999 0.105
Cedar Rapids – W. Burlington 0.999 0.003 0.360 0.675 0.999 0.003 1 494 5
23 W. Burlington – Clinton 0.999 0.001 -0.063 -0.757 0.225 0.025 0.999 0.001 0.999 - 0.999 0.088
Clinton – W. Burlington 0.999 0.000 -0.046 0.967 0.999 0.000 1 487 12
24 W. Burlington – Davenport -0.731 0.642 -0.033 -0.605 0.422 0.031 0.871 0.400 - - 0.996 0.063
Davenport – W. Burlington 0.678 0.692 0.105 0.926 0.996 0.053 3 486 11
25 W. Burlington – Eddyville -0.079 0.961 -0.034 -0.194 0.876 0.016 0.503 0.671 -0.960 -0.945 -0.918 0.050
Eddyville – W. Burlington -0.960 0.100 -0.945 0.067 -0.918 0.167 2 440 58
26 W. Burlington – Emmetsburg -0.580 0.245 -0.012 -0.201 0.796 0.025 -0.201 0.796 - - - 0.036
Emmetsburg – W. Burlington -0.081 0.912 0.438 0.529 0.438 0.529 139 322 42
27 W. Burlington – Keokuk -0.983 0.132 -0.021 -0.996 0.036 0.014 -0.947 0.256 - 0.996 - 0.035
Keokuk – W. Burlington -0.066 0.977 -0.733 0.571 0.590 0.717 10 459 31
36
d) North Carolina, the USA
Price pair Lower regime Middle regime Upper regime Total adjustment
Number of obs.
Dependent – indep. variable 𝝆𝟏 [Pvalue] Lower
Thresh.
𝝆𝟐 [Pvalue] Upper
Thresh.
𝝆𝟑 [Pvalue] Lower Middle Upper Band of
inaction
1 Candor – Creswell -0.804 0.018 -0.012 -0.468 0.338 0.029 -0.475 0.300 0.804 - - 0.040
Creswell – Candor -0.588 0.241 0.151 0.822 0.187 0.765 236 173 91
2 Cofield – Candor -0.761 0.097 -0.025 -0.773 0.116 0.019 -0.773 0.116 0.761 - - 0.044
Candor – Cofield -0.449 0.457 -0.494 0.439 -0.494 0.439 52 352 96
3 Cofield – Creswell -0.942 0.001 -0.043 -0.926 0.002 0.024 -0.761 0.067 0. 942 0. 926 0.761 0.067
Creswell – Cofield -0.798 0.070 -0.761 0.098 -0.257 0.719 4 409 87
4 Laurinburg – Candor -0.657 0.123 -0.032 -0.653 0.126 0.015 -0.713 0.061 - - 0.713 0.047
Candor – Laurinburg -0.454 0.361 -0.441 0.379 -0.547 0.216 47 360 93
5 Laurinburg – Cofield -0.831 0.078 -0.049 -0.774 0.119 0.003 -0.820 0.070 0.831 - 0.820 0.051
Cofield – Laurinburg -0.409 0.590 -0.443 0.531 -0.289 0.710 10 276 214
6 Laurinburg – Creswell -0.729 0.035 -0.034 -0.945 0.002 0.010 -0.729 0.035 0. 729 0.945 0.729 0.044
Creswell – Laurinburg -0.458 0.367 -0.886 0.026 -0.458 0.367 34 341 125
7 Laurinburg – Roaring River -0.622 0.227 -0.050 -0.622 0.227 0.037 -0.764 0.586 - - 0.999 0.088
Roaring River – Laurinburg -0.125 0.876 -0.125 0.876 0.999 0.034 2 495 3
8 Laurinburg – Statesville 0.178 0.843 -0.058 -0.730 0.166 0.016 0.178 0.843 0.910 - 0.910 0.074
Statesville – Laurinburg 0.910 0.025 0.049 0.956 0.910 0.025 10 428 62
9 Roaring River – Candor -1.000 0.000 -0.029 -0.871 0.106 0.024 -0.720 0.273 1.000 - - 0.053
Candor – Roaring River -0.965 0.099 -0.780 0.184 -0.277 0.753 11 443 46
10 Roaring River – Cofield -0.991 0.008 -0.144 -0.686 0.210 0.034 -0.686 0.210 0.991 - - 0.178
Cofield – Roaring River -0.103 0.942 -0.070 0.930 -0.070 0.930 3 474 23
11 Roaring River – Creswell -0.997 0.000 -0.125 -0.756 0.027 0.035 -0.756 0.027 0.997 0.756 0.756 0.160
Creswell – Roaring River -0.050 0.965 -0.618 0.132 -0.618 0.132 3 421 76
12 Roaring River – Statesville -0.991 0.026 -0.129 -0.477 0.531 0.028 -0.494 0.543 1.972 0.865 0.900 0.157
Statesville – Roaring River 0.981 0.045 0.865 0.043 0.900 0.036 3 478 19
13 Statesville – Candor 0.832 0.052 -0.051 -0.252 0.690 0.016 0.832 0.052 0.961 - 0.961 0.068
Candor – Statesville 0.961 0.001 0.449 0.429 0.961 0.001 8 398 94
14 Statesville – Cofield 0.204 0.747 -0.034 -0.151 0.828 0.027 -0.214 0.737 0.741 - - 0.061
Cofield – Statesville 0.741 0.062 0.591 0.241 0.539 0.281 28 420 52
15 Statesville – Creswell -0.879 0.005 -0.046 -0.914 0.001 0.004 -0.795 0.003 0.879 0.914 0.795 0.050
Creswell – Statesville 0.617 0.211 -0.710 0.102 -0.544 0.168 28 274 198
Note: Total adjustment in one regime is calculated as the sum of the absolute value of the respective regime-specific speed of adjustment parameters of the TVECM. The band of
inaction is given as the difference between the absolute value of the upper and lower threshold.Speed of adjustment coefficients which have opposite sign than threoretically
exected and are significant at least at 0.10 level are given in italic. Speed of adjustment parameters for Iowa and North Carolina in tabels c) and d) are adjusted from daily to by-
weekly frequencies in order to make coefficients comparable with the estimates for North Caucasus and West Siberia in tables a) and b). Source: Own estimations.
Table A5: Parameters of the long-run price equilibrium regression, the USA 2008/11
Dependent variable Independent variable distance Slope Intercept P-value
Arkansas Illinois 595 0.967 0.177 <0.001
Arkansas Iowa 475 0.974 0.161 <0.001
Arkansas Kansas 993 0.941 0.325 <0.001
Arkansas Minnesota 531 0.953 0.279 <0.001
Arkansas Missouri 393 0.929 0.406 <0.001
Arkansas Nebraska 581 0.979 0.109 0.094
Arkansas South Dakota 1144 0.894 0.593 <0.001
California Illinois 3288 0.764 1.968 <0.001
California Iowa 3084 0.776 1.93 <0.001
California Kansas 2356 0.758 2.019 <0.001
California Minnesota 3224 0.761 2.01 <0.001
California Missouri 2945 0.725 2.189 <0.001
California Nebraska 2675 0.784 1.863 <0.001
California South Dakota 2548 0.715 2.258 <0.001
Colorado Illinois 1720 0.982 0.045 0.369
Colorado Iowa 1273 0.997 -0.003 0.929
Colorado Kansas 494 0.979 0.088 <0.001
Colorado Minnesota 1482 0.978 0.103 <0.001
Colorado Missouri 974 0.929 0.346 <0.001
Colorado Nebraska 866 1.007 -0.082 0.032
Colorado South Dakota 901 0.919 0.418 <0.001
Oklahoma Illinois 1315 1.08 -0.437 <0.001
Oklahoma Iowa 1289 1.093 -0.473 <0.001
Oklahoma Kansas 220 1.078 -0.393 <0.001
Oklahoma Minnesota 1498 1.073 -0.361 <0.001
Oklahoma Missouri 789 1.02 -0.101 0.244
Oklahoma Nebraska 874 1.104 -0.558 <0.001
Oklahoma South Dakota 1073 1.009 -0.016 0.787
Oregon Illinois 3642 0.777 1.656 <0.001
Oregon Iowa 2836 0.788 1.619 <0.001
Oregon Kansas 2472 0.77 1.708 <0.001
Oregon Minnesota 2926 0.774 1.698 <0.001
Oregon Missouri 2895 0.735 1.889 <0.001
Oregon Nebraska 2660 0.796 1.554 <0.001
Oregon South Dakota 2245 0.728 1.945 <0.001
Texas Illinois 1226 0.951 0.26 <0.001
Texas Iowa 1487 0.963 0.223 <0.001
Texas Kansas 380 0.948 0.3 <0.001
Texas Minnesota 1695 0.945 0.322 <0.001
Texas Missouri 985 0.899 0.553 <0.001
Texas Nebraska 1032 0.972 0.148 <0.001
Texas South Dakota 1262 0.89 0.621 <0.001
Virginia Illinois 1349 0.922 0.49 <0.001
Virginia Iowa 1897 0.929 0.478 <0.001
Virginia Kansas 2356 0.909 0.577 <0.001
Virginia Minnesota 1833 0.908 0.588 <0.001
Virginia Missouri 1754 0.879 0.739 <0.001
Virginia Nebraska 2037 0.934 0.421 <0.001
Virginia South Dakota 2565 0.852 0.887 <0.001
Washington Illinois 3375 0.788 1.574 <0.001
Washington Iowa 2393 0.800 1.538 <0.001
Washington Kansas 2351 0.787 1.602 <0.001
Washington Minnesota 2482 0.785 1.618 <0.001
38
Washington Missouri 2628 0.746 1.811 <0.001
Washington Nebraska 2342 0.808 1.473 <0.001
Washington South Dakota 1801 0.739 1.870 <0.001
Wyoming Illinois 1782 0.974 0.088 0.079
Wyoming Iowa 1221 0.989 0.041 0.285
Wyoming Kansas 721 0.97 0.136 <0.001
Wyoming Minnesota 1310 0.97 0.147 <0.001
Wyoming Missouri 1033 0.921 0.387 <0.001
Wyoming Nebraska 800 0.999 -0.037 0.352
Wyoming South Dakota 653 0.912 0.458 <0.001
Note: All slope parameters are significant at a level lower than 1%. Weekly price series include 156 observations
recorded between 2008 and 2011 marketing years. Estimation of long-run price transmission coefficients was
preceeded by identification of linear or threshold cointegration (Johansen, 1988; Hansen and Seo 2006, Larsen,
2012). Tests confirmed existence of cointegration between regional market pairs in the USA. Results are available
from authors upon request.
Source: Own estimations.